Fix a bug discovered by the synthetic video.

Typo in calling ffmpeg.
Improve error handling.
2026-08-31 21:28:07 +02:00 · 2026-08-31 21:27:41 +02:00 · 2026-08-31 14:30:38 +02:00 · 2026-08-31 13:54:03 +02:00 · 2026-08-31 12:00:12 +02:00 · 2026-08-31 11:20:25 +02:00
21 changed files with 1031 additions and 909 deletions
@@ -6,46 +6,18 @@
 import argparse
 import logging
-import os.path
+from datetime import timedelta
 from datetime import datetime, timedelta
 from os import unlink
 from shutil import copyfile, move
 from sys import exit
 from functools import cmp_to_key
 from pathlib import Path
 import coloredlogs
 import hexdump
-from .tscut import (
+from tscut.exceptions import TSCutError
-    SupportedFormat,
+from tscut.models import ProcessingOptions
-    change_codec_private_data,
+from tscut.pipeline import process_recording
-    check_required_tools,
+from tscut.temporaries import TemporaryFiles
-    compare_time_interval,
+from tscut.tools.discovery import check_required_tools
-    concatenate_h264_parts,
+from tscut.tools.timeframe import compare_time_interval, parse_time_interval
    concatenate_h264_ts_parts,
    do_ocr,
    dump_codec_private_data,
    extract_all_streams,
    extract_mkv_part,
    extract_srt,
    extract_track_from_mkv,
    ffmpeg_convert,
    find_subtitles_tracks,
    get_avc_config_from_h264,
    get_codec_private_data_from_mkv,
    get_format,
    get_frame_rate,
    get_movie_duration,
    get_nearest_iframe,
    get_streams,
    get_tesseract_supported_lang,
    get_ts_frame,
    merge_mkvs,
    parse_codec_private,
    parse_time_interval,
    remove_video_tracks_from_mkv,
    remux_srt_subtitles,
 )
 logger = logging.getLogger(__name__)
@@ -88,7 +60,7 @@ def main() -> int:
    if args.coarse and args.threshold is not None:
        logger.error('--coarse and threshold arguments are exclusive.')
-        exit(-1)
+        return 2
    if (not args.coarse) and args.threshold is None:
        args.threshold = 0
@@ -108,7 +80,7 @@ def main() -> int:
        ts1, ts2 = parse_time_interval(interval)
        if ts1 is None or ts2 is None:
            logger.error("Illegal time interval: %s", interval)
-            exit(-1)
+            return 2
        parts.append((ts1,ts2))
    # Sort intervals
@@ -120,463 +92,32 @@ def main() -> int:
        ts1, ts2 = part
        if prevts > ts1:
            logger.error('Intervals are overlapping')
-            exit(-1)
+            return 2
        prevts = ts2
-    nb_parts = len(parts)
+    config = ProcessingOptions(
-    temporaries = []
+        input_file=Path(args.input_file),
-
+        output_file=Path(args.output_file),
-    basename = os.path.splitext(os.path.basename(args.input_file))[0]
+        parts=parts,
-    mp4filename = basename+'.mp4'
+        framerate = args.framerate,
-    mkvfilename = basename+'.mkv'
+        verbose=args.verbose,
        coarse=args.coarse,
        threshold=args.threshold,
        keep_temporaries=args.keep,
        subtitles_ocr=args.srt,
        dump_memory=args.dump,
        tools_paths = paths,
        all_optional_tools = all_optional_tools)
    logger.debug("Configuration: %s", config)
    temporaries = TemporaryFiles(config.keep_temporaries)
    try:
-        input_file = open(args.input_file, mode='rb')
+        process_recording(config, temporaries)
-        logger.debug("Type of input file: %s", type(input_file))
+    except TSCutError as exc:
-    except OSError:
+        logger.error("%s", exc)
-        logger.error("Impossible to open %s", args.input_file)
+        return 1
-        exit(-1)
+    finally:
    format_of_file = get_format(paths['ffprobe'], input_file)
    if format_of_file is None:
        exit(-1)
    duration = timedelta(seconds=float(format_of_file['duration']))
    logger.info("Durée de l'enregistrement: %s", duration)
    if args.framerate is None:
        framerate = get_frame_rate(paths['ffprobe'], input_file)
        if framerate is None:
            logger.error('Impossible to estimate frame rate !')
            exit(-1)
    else:
        framerate = args.framerate
    logger.info('Frame rate: %.1f fps', framerate)
    found = False
    for f in SupportedFormat:
        if 'format_name' in format_of_file:
            if format_of_file['format_name'] == str(f):
                found = True
                format_of_file = f
                break
    if not found:
        logger.error('Unsupported format of file')
    if format_of_file == SupportedFormat.TS:
        logger.info("Converting TS to MP4 (to fix timestamps).")
        try:
            with open(mp4filename, 'wb+') as mp4:
                ffmpeg_convert(paths['ffmpeg'], paths['ffprobe'], input_file, 'mpegts', mp4, 'mp4',
                              duration)
                temporaries.append(mp4)
                logger.info("Converting MP4 to MKV.")
                try:
                    mkv = open(mkvfilename, 'wb+')
                except OSError:
                    logger.error('')
                ffmpeg_convert(paths['ffmpeg'], paths['ffprobe'], mp4, 'mp4', mkv, 'matroska',
                              duration)
                if nb_parts > 0:
                    temporaries.append(mkv)
        except OSError:
            logger.error('')
    elif format_of_file == SupportedFormat.MP4:
        logger.info("Converting MP4 to MKV")
        try:
            mkv = open(mkvfilename, 'wb+')
        except OSError:
            logger.error('')
        ffmpeg_convert(paths['ffmpeg'], paths['ffprobe'], input_file, 'mp4', mkv, 'matroska',
                      duration)
        if nb_parts > 0:
            temporaries.append(mkv)
    else:
        logger.info("Already in MKV")
        mkv = input_file
    streams = get_streams(paths['ffprobe'], mkv)
    logger.debug('Streams: %s', streams)
    main_video = None
    nb_videos = 0
    for stream in streams:
        if stream['codec_type'] == 'video':
            if stream['disposition']['default'] == 1:
                main_video = stream
                width = stream['width']
                height = stream['height']
                break
            nb_videos+=1
            if nb_videos == 1:
                main_video = stream
                width = stream['width']
                height = stream['height']
            else:
                main_video = None
    if main_video is None:
        logger.error('Impossible to find main video stream.')
        exit(-1)
    # We retrieve the main private codec data
    _, main_codec_private_data = get_codec_private_data_from_mkv(mkvinfo_path=paths['mkvinfo'],
                                                                 input_file=mkv)
    logger.debug('Main video stream has following private data: %s',
                 hexdump.dump(main_codec_private_data, sep=':'))
    # We parse them
    main_avc_config = parse_codec_private(main_codec_private_data)
    logger.debug('AVC configuration: %s', main_avc_config)
    # We check if the parse and dump operations are idempotent.
    private_data = dump_codec_private_data(main_avc_config)
    logger.debug('Redump AVC configuration: %s', hexdump.dump(private_data, sep=':'))
    # In rare occasion, the PPS has trailing zeroes that do not seem to be related to useful data
    # but they differ from the private data we generate that do not contain them.
    # In that case we try to redecode our own private data to see if both AVC configurations are
    # the same.
    if main_codec_private_data != private_data:
        logger.warning('Difference detected in bitstream !!')
        iso_avc_config = parse_codec_private(private_data)
        logger.debug('Reread AVC configuration: %s', iso_avc_config)
        # If there exists a difference between our own reconstructed AVC configuration and the
        # original one, we abandon
        if iso_avc_config != main_avc_config:
            logger.error('AVC configurations are different: %s\n%s\n', main_avc_config,
                         iso_avc_config)
            exit(-1)
    # Pour chaque portion
    partnum = 0
    mkvparts = []
    h264parts = []
    h264_ts = []
    checks = []
    pos = timedelta()
    other_avc_configs = []
    for ts1, ts2 in parts:
        # TODO: translate comment in english
        # Trouver l'estampille de la trame 'I' la plus proche (mais postérieure) au début
        # de la portion.
        # Trouver l'estampille de la trame 'I' la plus proche (mais antérieure) à la fin
        # de la portion.
        # On a alors
        # debut  -----    trame  --------- trame  ---------  fin fin+1
        #  'B/P'  'B/P'*   'I'               'I'   'B/P'*   'B/P' 'I/B/P'
        # Si la trame de début est déjà 'I', il n'y a rien à faire.
        # Sinon on extrait les trames 'B' ou 'P' depuis le début jusqu'à la trame 'I' non incluse.
        # Si la trame de fin précède une trame I, on n'a rien à faire.
        # Sinon on extrait toutes les trames depuis la dernière trame I jusqu'à la trame de fin.
        partnum = partnum + 1
        # Get the nearest I-frame whose timestamp is greater or equal to the beginning.
        head_frames = get_nearest_iframe(paths['ffprobe'], mkv, ts1, before=False)
        if head_frames is None:
            logger.error('Impossible to retrieve I-frame')
            exit(-1)
        # Get the nearest I-frame whose timestamp ...
        # TODO: wrong here ...
        tail_frames = get_nearest_iframe(paths['ffprobe'], mkv, ts2, before=True)
        if tail_frames is None:
            logger.error('Impossible to retrieve I-frame')
            exit(-1)
        nb_head_frames, head_iframe = head_frames
        nb_tail_frames, tail_iframe = tail_frames
        logger.info("Found %d frames between beginning of current part and first I-frame",
                    nb_head_frames)
        logger.info("Found %d frames between last I-frame and end of current part",
                    nb_tail_frames)
        head_iframe_ts = get_ts_frame(head_iframe)
        if head_iframe_ts is None:
            exit(-1)
        tail_iframe_ts = get_ts_frame(tail_iframe)
        if tail_iframe_ts is None:
            exit(-1)
        checks.append(pos+head_iframe_ts-ts1)
        subparts = []
        # TODO: separate pipeline processing between coarse and not fine grain options.
        # if args.coarse:
        #     do_coarse_processing(ffmpeg=paths['ffmpeg'], ffprobe=paths['ffprobe'], input_file=mkv,
        #                        begin=ts1, end=head_iframe_ts, nb_frames=nb_head_frames-1,
        #                        frameRate=frameRate, files_prefix='part-%d-head' % (partnum),
        #                        streams=streams, width=width, height=height,
        #                        temporaries=temporaries, dump_mem_fd=args.dump)
        # else:
        #     doFineGrainProcessing(ffmpeg=paths['ffmpeg'], ffprobe=paths['ffprobe'],
        #                           input_file=mkv, begin=ts1, end=head_iframe_ts,
        #                           nb_frames=nb_head_frames-1, frameRate=frameRate,
        #                           files_prefix='part-%d-head' % (partnum), streams=streams,
        #                           width=width, height=height, temporaries=temporaries,
        #                           dump_mem_fd=args.dump)
        if (not args.coarse) and (nb_head_frames > args.threshold):
            # We extract all frames between the beginning upto the frame that immediately preceeds
            # the I-frame.
            h264_head, h264_head_ts, mkv_head = extract_all_streams(ffmpeg_path=paths['ffmpeg'],
                                                              ffprobe_path=paths['ffprobe'],
                                                              input_file=mkv, begin=ts1,
                                                              end=head_iframe_ts,
                                                              nb_frames=nb_head_frames-1,
                                                              framerate=framerate,
                                                              files_prefix=f'part-{partnum:d}-head',
                                                              streams=streams, width=width,
                                                              height=height,
                                                              temporaries=temporaries,
                                                              dump_mem_fd=args.dump)
            # If we are not at an exact boundary:
            if mkv_head is not None:
                subparts.append(mkv_head)
            if h264_head is not None:
                avcconfig = get_avc_config_from_h264(h264_head)
                other_avc_configs.append(avcconfig)
                h264parts.append(h264_head)
            if h264_head_ts is not None:
                h264_ts.append(h264_head_ts)
        # Creating MKV file that corresponds to current part between I-frames
        # Internal video with all streams (video, audio and subtitles)
        internal_mkv_name = f'part-{partnum:d}-internal.mkv'
        # Internal video stream as a raw H264 stream
        internal_h264_name = f'part-{partnum:d}-internal.h264'
        # Internal video timestamps
        internal_h264_ts_name = f'part-{partnum:d}-internal-ts.txt'
        # Internal video with only audio and subtitles streams
        internal_novideo_mkv_name = f'part-{partnum:d}-internal-novideo.mkv'
        try:
            internal_mkv = open(internal_mkv_name, 'wb+')
        except OSError:
            logger.error('Impossible to create file: %s', internal_mkv_name)
            exit(-1)
        try:
            internal_novideo_mkv = open(internal_novideo_mkv_name, 'wb+')
        except OSError:
            logger.error('Impossible to create file: %s', internal_novideo_mkv_name)
            exit(-1)
        try:
            internal_h264 = open(internal_h264_name, 'wb+')
        except OSError:
            logger.error('Impossible to create file: %s', internal_h264_name)
            exit(-1)
        try:
            internal_h264_ts = open(internal_h264_ts_name, 'w+', encoding='utf8')
        except OSError:
            logger.error('Impossible to create file: %s', internal_h264_ts_name)
            exit(-1)
        # logger.info('Merge header, middle and trailer subpart into: %s' % internal_mkv_name)
        # Extract internal part of MKV
        extract_mkv_part(mkvmerge_path=paths['mkvmerge'], input_file=mkv, output_file=internal_mkv,
                       begin=head_iframe_ts, end=tail_iframe_ts)
        # Extract video stream of internal part as a raw H264 and its timestamps.
        logger.info('Extract video track as raw H264 file.')
        extract_track_from_mkv(mkvextract_path=paths['mkvextract'], input_file=internal_mkv,
                               index=0, output_file=internal_h264, timestamps=internal_h264_ts)
        # Remove video track from internal part of MKV
        logger.info('Remove video track from %s', internal_mkv_name)
        remove_video_tracks_from_mkv(mkvmerge_path=paths['mkvmerge'], input_file=internal_mkv,
                                 output_file=internal_novideo_mkv)
        temporaries.append(internal_mkv)
        temporaries.append(internal_h264)
        temporaries.append(internal_h264_ts)
        temporaries.append(internal_novideo_mkv)
        h264parts.append(internal_h264)
        h264_ts.append(internal_h264_ts)
        subparts.append(internal_novideo_mkv)
        if (not args.coarse) and (nb_tail_frames > args.threshold):
            # We extract all frames between the I-frame (including it) upto the end.
            h264_tail, h264_tail_ts, mkv_tail = extract_all_streams(ffmpeg_path=paths['ffmpeg'],
                                                              ffprobe_path=paths['ffprobe'],
                                                              input_file=mkv, begin=tail_iframe_ts,
                                                              end=ts2, nb_frames=nb_tail_frames,
                                                              framerate=framerate,
                                                              files_prefix=f'part-{partnum:d}-tail',
                                                              streams=streams,
                                                              width=width, height=height,
                                                              temporaries=temporaries,
                                                              dump_mem_fd=args.dump)
            if mkv_tail is not None:
                subparts.append(mkv_tail)
            if h264_tail is not None:
                avcconfig = get_avc_config_from_h264(h264_tail)
                other_avc_configs.append(avcconfig)
                h264parts.append(h264_tail)
            if h264_tail_ts is not None:
                h264_ts.append(h264_tail_ts)
        logger.info('Merging MKV: %s', subparts)
        part = merge_mkvs(mkvmerge_path=paths['mkvmerge'], inputs=subparts,
                         output_name=f'part-{partnum:d}.mkv', concatenate=True)
        mkvparts.append(part)
        temporaries.append(part)
        pos = pos+tail_iframe_ts-ts1
    # We need to check the end also
    checks.append(pos)
    # When using coarse option there is a single AVC configuration.
    for avc_config in other_avc_configs:
        main_avc_config.merge(avc_config)
        logger.debug('Merged AVC configuration: %s', main_avc_config)
    nb_mkv_parts = len(mkvparts)
    if nb_mkv_parts > 0:
        try:
            full_h264 = open(f'{basename}-full.h264', 'wb+')
        except OSError:
            logger.error('Impossible to create file full H264 stream.')
            exit(-1)
        logger.info('Merging all H264 tracks')
        concatenate_h264_parts(h264parts=h264parts, output=full_h264)
        temporaries.append(full_h264)
        try:
            full_h264_ts = open(f'{basename}-ts.txt', 'w+', encoding='utf8')
        except OSError:
            logger.error('Impossible to create file containing all video timestamps.')
            exit(-1)
        logger.info('Merging H264 timestamps')
        concatenate_h264_ts_parts(h264_ts_parts=h264_ts, output=full_h264_ts)
        temporaries.append(full_h264_ts)
    final_novideo_name = f'{basename}-novideo.mkv'
    final_with_video_name = f'{basename}-video.mkv'
    if nb_mkv_parts > 1:
        logger.info('Merging all audio and subtitles parts: %s', mkvparts)
        merge_mkvs(mkvmerge_path=paths['mkvmerge'], inputs=mkvparts, output_name=final_novideo_name,
                  concatenate=True)
    elif nb_mkv_parts == 1:
        copyfile('part-1.mkv', final_novideo_name)
    else:
        logger.info("Nothing else to do.")
        copyfile(mkvfilename, final_with_video_name)
    if nb_mkv_parts >=1 :
        try:
            final_novideo = open(final_novideo_name, 'rb')
        except OSError:
            logger.error('Impossible to open file: %s.', final_novideo_name)
            exit(-1)
        temporaries.append(final_novideo)
        full_h264_ts.seek(0)
        logger.info('Merging final video track and all other tracks together')
        final_with_video = merge_mkvs(mkvmerge_path=paths['mkvmerge'], inputs=[full_h264,
                                                                               final_novideo],
                                   output_name=final_with_video_name, concatenate=False,
                                   timestamps={0: full_h264_ts})
        final_codec_private_data = dump_codec_private_data(main_avc_config)
        logger.debug('Final codec private data: %s', hexdump.dump(final_codec_private_data,
                                                                  sep=':'))
        logger.info('Changing codec private data with the new one.')
        change_codec_private_data(paths['mkvinfo'], final_with_video, final_codec_private_data)
    if args.srt:
        if not all_optional_tools:
            logger.warning("Missing tools for extracting subtitles.")
            move(final_with_video_name, args.output_file)
        else:
            # Final cut is not any more the final step.
            temporaries.append(final_with_video)
            duration = get_movie_duration(paths['ffprobe'], final_with_video)
            supported_langs = get_tesseract_supported_lang(paths['tesseract'])
            logger.info('Supported lang: %s', supported_langs)
            logger.info('Find subtitles tracks and language.')
            subtitles = find_subtitles_tracks(paths['ffprobe'], final_with_video)
            logger.info(subtitles)
            sts = {}
            for subtitle in subtitles:
                index = subtitle['index']
                if 'tags' in subtitle:
                    if 'language' in subtitle['tags']:
                        lang = subtitle['tags']['language']
                        if lang in sts:
                            sts[lang].append(index)
                        else:
                            sts[lang] = [index]
                    else:
                        logger.error("Dropping subtitle: %s because it is missing language\
                            indication", subtitle)
                else:
                    logger.error("Dropping subtitle: %s because it is missing language indication",
                                 subtitle)
            logger.info(sts)
            if len(sts) > 0:
                logger.info('Supported languages: %s', supported_langs)
                list_of_subtitles = extract_srt(paths['mkvextract'], final_with_video_name, sts,
                                             supported_langs)
                logger.info(list_of_subtitles)
                for idx_name, sub_name, _, _ in list_of_subtitles:
                    try:
                        idx = open(idx_name,'rb')
                    except OSError:
                        logger.error("Impossible to open %s.", idx_name)
                        exit(-1)
                    try:
                        sub = open(sub_name,'rb')
                    except OSError:
                        logger.error("Impossible to open %s.", sub_name)
                        exit(-1)
                    temporaries.append(idx)
                    temporaries.append(sub)
                ocr = do_ocr(paths['vobsubocr'], list_of_subtitles, duration, temporaries,
                             args.dump)
                logger.info(ocr)
                # Remux SRT subtitles
                remux_srt_subtitles(paths['mkvmerge'], final_with_video, args.output_file, ocr)
            else:
                copyfile(final_with_video_name, args.output_file)
    else:
        move(final_with_video_name, args.output_file)
    if not args.keep:
        logger.info("Cleaning temporary files")
-        for f in temporaries:
+        temporaries.cleanup()
            path = os.path.realpath(f.name)
            logger.info("Removing: %s", path)
            f.close()
            unlink(path)
    d = datetime(1,1,1)
    for c in checks:
        logger.info("Please check cut smoothness at %s", (c+d).strftime("%H:%M:%S"))
    return 0
@@ -0,0 +1,30 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 import logging
 from typing import IO
 from tscut.exceptions import TemporaryFileError
 from tscut.tools.mkvtoolnix import extract_mkv_part
 logger = logging.getLogger(__name__)
 # TODO: finish this procedure
 def do_coarse_processing(ffmpeg_path:str, ffprobe_path:str, mkvmerge_path:str,
                         input_file: IO[bytes], begin, end, nb_frames, framerate,
                         files_prefix, streams, width, height, temporaries, dump_mem_fd) -> None:
    # Internal video with all streams (video, audio and subtitles)
    internal_mkv_name = f'{files_prefix}.mkv'
    try:
        internal_mkv = open(internal_mkv_name, 'wb+')
    except OSError as e:
        raise TemporaryFileError(f"Impossible to create {internal_mkv_name}") from e
    # Extract internal part of MKV
    extract_mkv_part(mkvmerge_path=mkvmerge_path, input_file=input_file, output_file=internal_mkv,
                     begin=begin, end=end)
    temporaries.add(internal_mkv)
@@ -0,0 +1,365 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 import logging
 from datetime import timedelta
 from os import (
    SEEK_SET,
    fstat,
    lseek,
    read,
    write,
 )
 from shutil import copyfile
 from typing import BinaryIO, Sequence, TextIO
 import hexdump
 from tqdm import tqdm
 from typeguard import typechecked
 from tscut.exceptions import InvalidMediaError, TemporaryFileError
 from tscut.h264.avc import get_avc_config_from_h264
 from tscut.matroska.codec import dump_codec_private_data
 from tscut.models import CutResult, PreparedMedia, ProcessingOptions
 from tscut.temporaries import TemporaryFiles
 from tscut.tools.ffmpeg import extract_all_streams
 from tscut.tools.ffprobe import (
    get_nearest_iframe,
 )
 from tscut.tools.mkvtoolnix import (
    change_codec_private_data,
    extract_mkv_part,
    extract_track_from_mkv,
    merge_mkvs,
    remove_video_tracks_from_mkv,
 )
 from tscut.tools.timeframe import get_ts_frame
 logger = logging.getLogger(__name__)
@typechecked
 def concatenate_h264_parts(h264parts: Sequence[BinaryIO], output: BinaryIO) -> None:
    logger = logging.getLogger(__name__)
    total_length = 0
    for h264 in h264parts:
        fd = h264.fileno()
        total_length += fstat(fd).st_size
    logger.info('Total length: %d', total_length)
    outfd = output.fileno()
    lseek(outfd, 0, SEEK_SET)
    pb = tqdm(total=total_length, unit='bytes', desc='Concatenation')
    for h264 in h264parts:
        fd = h264.fileno()
        lseek(fd, 0, SEEK_SET)
        while True:
            buf = read(fd, 1000000)
            if buf is None or len(buf) == 0:
                break
            pos = 0
            while pos < len(buf):
                nb_bytes = write(outfd, buf[pos:])
                pb.update(nb_bytes)
                pos += nb_bytes
 def concatenate_h264_ts_parts(h264_ts_parts: Sequence[TextIO], output: TextIO) -> None:
    logger = logging.getLogger(__name__)
    header = '# timestamp format v2\n'
    output.write(header)
    last = 0.
    first = True
    for part in h264_ts_parts:
        if first:
            offset = last
        else:
            # TODO: take framerate into account
            offset = last + 40
        logger.debug('Parsing file: %s. Offset=%d', part, offset)
        isheader = part.readline()
        if (not isheader) or (isheader != header):
            raise InvalidMediaError(f"Impossible to find a valid header: {isheader}")
        while True:
            line = part.readline()
            if not line:
                break
            ts = offset + float(line)
            last = max(last,ts)
            output.write(f'{ts:f}\n')
        if first:
            first = False
 def cut_recording(media: PreparedMedia, options: ProcessingOptions,
                  temporaries: TemporaryFiles) -> CutResult:
    # Pour chaque portion
    partnum = 0
    mkvparts = []
    h264parts = []
    h264_ts = []
    checks = []
    pos = timedelta()
    other_avc_configs = []
    mkvfilename = media.basename+'.mkv'
    for ts1, ts2 in options.parts:
        # TODO: translate comment in english
        # Trouver l'estampille de la trame 'I' la plus proche (mais postérieure) au début
        # de la portion.
        # Trouver l'estampille de la trame 'I' la plus proche (mais antérieure) à la fin
        # de la portion.
        # On a alors
        # debut  -----    trame  --------- trame  ---------  fin fin+1
        #  'B/P'  'B/P'*   'I'               'I'   'B/P'*   'B/P' 'I/B/P'
        # Si la trame de début est déjà 'I', il n'y a rien à faire.
        # Sinon on extrait les trames 'B' ou 'P' depuis le début jusqu'à la trame 'I' non incluse.
        # Si la trame de fin précède une trame I, on n'a rien à faire.
        # Sinon on extrait toutes les trames depuis la dernière trame I jusqu'à la trame de fin.
        partnum = partnum + 1
        # Get the nearest I-frame whose timestamp is greater or equal to the beginning.
        head_frames = get_nearest_iframe(options.tools_paths['ffprobe'], media.movie, ts1,
                                         before=False)
        if head_frames is None:
            raise InvalidMediaError("Impossible to retrieve first I-frame")
        # Get the nearest I-frame whose timestamp ...
        # TODO: wrong here ...
        tail_frames = get_nearest_iframe(options.tools_paths['ffprobe'], media.movie,
                                         ts2, before=True)
        if tail_frames is None:
            raise InvalidMediaError("Impossible to retrieve last I-frame")
        nb_head_frames, head_iframe = head_frames
        nb_tail_frames, tail_iframe = tail_frames
        logger.info("Found %d frames between beginning of current part and first I-frame",
                    nb_head_frames)
        logger.info("Found %d frames between last I-frame and end of current part",
                    nb_tail_frames)
        try:
            head_iframe_ts = get_ts_frame(head_iframe)
        except InvalidMediaError:
            raise InvalidMediaError("Impossible to retrieve timestamp of first i-frame.")
        try:
            tail_iframe_ts = get_ts_frame(tail_iframe)
        except InvalidMediaError:
            raise InvalidMediaError("Impossible to retrieve timestamp of final i-frame.")
        assert head_iframe_ts <= tail_iframe_ts
        checks.append(pos+head_iframe_ts-ts1)
        subparts = []
        # TODO: separate pipeline processing between coarse and not fine grain options.
        # if args.coarse:
        #     do_coarse_processing(ffmpeg=paths['ffmpeg'], ffprobe=paths['ffprobe'], input_file=mkv,
        #                        begin=ts1, end=head_iframe_ts, nb_frames=nb_head_frames-1,
        #                        frameRate=frameRate, files_prefix='part-%d-head' % (partnum),
        #                        streams=streams, width=width, height=height,
        #                        temporaries=temporaries, dump_mem_fd=args.dump)
        # else:
        #     doFineGrainProcessing(ffmpeg=paths['ffmpeg'], ffprobe=paths['ffprobe'],
        #                           input_file=mkv, begin=ts1, end=head_iframe_ts,
        #                           nb_frames=nb_head_frames-1, frameRate=frameRate,
        #                           files_prefix='part-%d-head' % (partnum), streams=streams,
        #                           width=width, height=height, temporaries=temporaries,
        #                           dump_mem_fd=args.dump)
        if (not options.coarse) and (nb_head_frames > options.threshold):
            # We extract all frames between the beginning upto the frame that immediately preceeds
            # the I-frame.
            h264_head, h264_head_ts, mkv_head = extract_all_streams(
                                                        ffmpeg_path=options.tools_paths['ffmpeg'],
                                                        ffprobe_path=options.tools_paths['ffprobe'],
                                                        input_file=media.movie, begin=ts1,
                                                        end=head_iframe_ts,
                                                        nb_frames=nb_head_frames-1,
                                                        framerate=media.framerate,
                                                        files_prefix=f'part-{partnum:d}-head',
                                                        streams=media.streams, width=media.width,
                                                        height=media.height,
                                                        temporaries=temporaries,
                                                        dump_mem_fd=options.dump_memory)
            # If we are not at an exact boundary:
            if mkv_head is not None:
                subparts.append(mkv_head)
            if h264_head is not None:
                avcconfig = get_avc_config_from_h264(h264_head)
                other_avc_configs.append(avcconfig)
                h264parts.append(h264_head)
            if h264_head_ts is not None:
                h264_ts.append(h264_head_ts)
        if head_iframe_ts < tail_iframe_ts:
            # Creating MKV file that corresponds to current part between I-frames
            # Internal video with all streams (video, audio and subtitles)
            internal_mkv_name = f'part-{partnum:d}-internal.mkv'
            # Internal video stream as a raw H264 stream
            internal_h264_name = f'part-{partnum:d}-internal.h264'
            # Internal video timestamps
            internal_h264_ts_name = f'part-{partnum:d}-internal-ts.txt'
            # Internal video with only audio and subtitles streams
            internal_novideo_mkv_name = f'part-{partnum:d}-internal-novideo.mkv'
            try:
                internal_mkv = open(internal_mkv_name, 'wb+')
            except OSError as e:
                raise TemporaryFileError(f"Impossible to create file: {internal_mkv_name}") from e
            try:
                internal_novideo_mkv = open(internal_novideo_mkv_name, 'wb+')
            except OSError as e:
                raise TemporaryFileError(f"Impossible to create file: \
                    {internal_novideo_mkv_name}") from e
            try:
                internal_h264 = open(internal_h264_name, 'wb+')
            except OSError as e:
                raise TemporaryFileError(f"Impossible to create file: {internal_h264_name}") from e
            try:
                internal_h264_ts = open(internal_h264_ts_name, 'w+', encoding='utf8')
            except OSError as e:
                raise TemporaryFileError(f"Impossible to create file: {internal_h264_ts_name}") from e
            # logger.info('Merge header, middle and trailer subpart into: %s' % internal_mkv_name)
            # Extract internal part of MKV
            extract_mkv_part(mkvmerge_path=options.tools_paths['mkvmerge'], input_file=media.movie,
                            output_file=internal_mkv, begin=head_iframe_ts, end=tail_iframe_ts)
            # Extract video stream of internal part as a raw H264 and its timestamps.
            logger.info('Extract video track as raw H264 file.')
            extract_track_from_mkv(mkvextract_path=options.tools_paths['mkvextract'],
                                input_file=internal_mkv, index=0, output_file=internal_h264,
                                timestamps=internal_h264_ts)
            # Remove video track from internal part of MKV
            logger.info('Remove video track from %s', internal_mkv_name)
            remove_video_tracks_from_mkv(mkvmerge_path=options.tools_paths['mkvmerge'],
                                        input_file=internal_mkv, output_file=internal_novideo_mkv)
            temporaries.add(internal_mkv)
            temporaries.add(internal_h264)
            temporaries.add(internal_h264_ts)
            temporaries.add(internal_novideo_mkv)
            h264parts.append(internal_h264)
            h264_ts.append(internal_h264_ts)
            subparts.append(internal_novideo_mkv)
        if (not options.coarse) and (nb_tail_frames > options.threshold):
            # We extract all frames between the I-frame (including it) upto the end.
            h264_tail, h264_tail_ts, mkv_tail = extract_all_streams(
                                ffmpeg_path=options.tools_paths['ffmpeg'],
                                ffprobe_path=options.tools_paths['ffprobe'],
                                input_file=media.movie, begin=tail_iframe_ts,
                                end=ts2, nb_frames=nb_tail_frames,
                                framerate=media.framerate,
                                files_prefix=f'part-{partnum:d}-tail',
                                streams=media.streams,
                                width=media.width, height=media.height,
                                temporaries=temporaries,
                                dump_mem_fd=options.dump_memory)
            if mkv_tail is not None:
                subparts.append(mkv_tail)
            if h264_tail is not None:
                avcconfig = get_avc_config_from_h264(h264_tail)
                other_avc_configs.append(avcconfig)
                h264parts.append(h264_tail)
            if h264_tail_ts is not None:
                h264_ts.append(h264_tail_ts)
        logger.info('Merging MKV: %s', subparts)
        part = merge_mkvs(mkvmerge_path=options.tools_paths['mkvmerge'], inputs=subparts,
                         output_name=f'part-{partnum:d}.mkv', concatenate=True)
        mkvparts.append(part)
        temporaries.add(part)
        pos = pos+tail_iframe_ts-ts1
    # We need to check the end also
    checks.append(pos)
    # When using coarse option there is a single AVC configuration.
    for avc_config in other_avc_configs:
        media.avc_config.merge(avc_config)
        logger.debug('Merged AVC configuration: %s', media.avc_config)
    nb_mkv_parts = len(mkvparts)
    if nb_mkv_parts > 0:
        try:
            full_h264 = open(f'{media.basename}-full.h264', 'wb+')
        except OSError as e:
            raise TemporaryFileError("Impossible to create file full H264 stream.") from e
        logger.info('Merging all H264 tracks')
        concatenate_h264_parts(h264parts=h264parts, output=full_h264)
        temporaries.add(full_h264)
        try:
            full_h264_ts = open(f'{media.basename}-ts.txt', 'w+', encoding='utf8')
        except OSError as e:
            raise TemporaryFileError("Impossible to create file containing all video timestamps.")\
            from e
        logger.info('Merging H264 timestamps')
        concatenate_h264_ts_parts(h264_ts_parts=h264_ts, output=full_h264_ts)
        temporaries.add(full_h264_ts)
    final_novideo_name = f'{media.basename}-novideo.mkv'
    final_with_video_name = f'{media.basename}-video.mkv'
    if nb_mkv_parts > 1:
        logger.info('Merging all audio and subtitles parts: %s', mkvparts)
        merge_mkvs(mkvmerge_path=options.tools_paths['mkvmerge'], inputs=mkvparts,
                   output_name=final_novideo_name, concatenate=True)
    elif nb_mkv_parts == 1:
        copyfile('part-1.mkv', final_novideo_name)
    else:
        logger.info("Nothing else to do.")
        copyfile(mkvfilename, final_with_video_name)
    if nb_mkv_parts >=1 :
        try:
            final_novideo = open(final_novideo_name, 'rb')
        except OSError as e:
            raise TemporaryFileError(f"Impossible to open file: {final_novideo_name}") from e
        temporaries.add(final_novideo)
        full_h264_ts.seek(0)
        logger.info('Merging final video track and all other tracks together')
        final_with_video = merge_mkvs(mkvmerge_path=options.tools_paths['mkvmerge'],
                                        inputs=[full_h264, final_novideo],
                                        output_name=final_with_video_name, concatenate=False,
                                        timestamps={0: full_h264_ts})
        final_codec_private_data = dump_codec_private_data(media.avc_config)
        logger.debug('Final codec private data: %s', hexdump.dump(final_codec_private_data,
                                                                  sep=':'))
        logger.info('Changing codec private data with the new one.')
        change_codec_private_data(options.tools_paths['mkvinfo'], final_with_video,
                                  final_codec_private_data)
    return CutResult(
        filename = final_with_video_name,
        movie = final_with_video,
        check_positions = checks
    )
@@ -0,0 +1,26 @@
 Useful SPS/PPS discussion:
 1.https://copyprogramming.com/howto/including-sps-and-pps-in-a-raw-h264-track
 2. https://gitlab.com/mbunkus/mkvtoolnix/-/issues/2390
 Strategy: a possible way of handling multiple SPS/PPS gracefully.
 Encode each head and trailer with FFMPEG using only I-frame (to be sure the NAL unit will never refer to another image).
 Encode using an different SPS-ID all of them (using sps-id parameter of libx264 library, e.g  1 instead of 0).
 For the video track produce only a raw H264 file and a file containing timestamps of the different frames.
 For the rest of the tracks (audio, subtitles) produce directly a MKV (this is already done).
 Concatenate all raw H264 in a giant one (like cat), and the same for timestamps of video frames (to keep sound and video synchronized).
 Then use mkvmerge to remux the H264 track and the rest of tracks.
 MKVmerge "concatenate" subcommand is able to concatenate different SPS/PPS data into a bigger Private Codec Data.
 However, this is proved to be not reliable.
 Sometimes it results in a AVC context containing a single SPS/PPS.
 So we have to rely on a manual parsing of the H264 AVC context of original movie and the ones produced for headers and trailers, and then merging them into a bigger AVC context.
 Then finally, change the Private Codec Data in the final MKV.
 Extract SPS/PPS:
 1. https://gitlab.com/mbunkus/mkvtoolnix/-/issues/2390
 ```bash
 ffmpeg -i <InputFile (before concatenation)> -c:v copy -an -sn -bsf:v trace_headers -t 0.01 -report -loglevel 0 -f null -
 ``
@@ -0,0 +1,24 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 class TSCutError(Exception):
    """Base exception for tscut."""
 class MissingToolError(TSCutError):
    """Raised when a required external tool is unavailable."""
 class ExternalToolError(TSCutError):
    """Raised when an external tool fails."""
 class InvalidMediaError(TSCutError):
    """Raised when input media cannot be processed."""
 class TemporaryFileError(TSCutError):
    """Raised when the manipulation of temporary file goes wrong"""
 class UnimplementedFeatureError(TSCutError):
    """Raised when a rare feature is encountered and not yet implemented"""
@@ -8,26 +8,24 @@ from dataclasses import dataclass, field
 from math import floor
 from typing import IO
 from typeguard import typechecked
 import hexdump
 from typeguard import typechecked
 from tscut.h264.bitstream import (
    rbsp_to_sodb,
    read_bit,
    read_bits,
    read_byte,
    read_long,
    read_word,
    sodb_to_rbsp,
    write_bits,
    write_byte,
    write_word,
    sodb_to_rbsp,
    rbsp_to_sodb
 )
 from tscut.h264.parameters import (
    SPS,
    PPS,
    SPS,
 )
 logger = logging.getLogger(__name__)
@@ -243,7 +241,7 @@ class AVCDecoderConfiguration:
        # TODO: do the same with extended SPS !
@typechecked
-def parse_codec_private(codec_private_data: bytes) -> AVCDecoderConfiguration:
+def parse_codec_private(codec_private_data: bytes | bytearray) -> AVCDecoderConfiguration:
    if codec_private_data[0] != 0x63:
        raise ValueError(f'Matroska header is wrong: {codec_private_data[0]:x}')
    if codec_private_data[1] != 0xA2:
@@ -10,14 +10,15 @@
 import logging
 from math import floor, log
 from typeguard import typechecked
 import hexdump
 from typeguard import typechecked
 from tscut.exceptions import UnimplementedFeatureError
 logger = logging.getLogger(__name__)
@typechecked
-def read_bit(buf:bytes, bit_position: int) -> tuple[int, int]:
+def read_bit(buf:bytes|bytearray, bit_position: int) -> tuple[int, int]:
    """
    Read a single bit from a byte buffer.
@@ -40,7 +41,7 @@ def read_bit(buf:bytes, bit_position: int) -> tuple[int, int]:
    return bit_position+1, bit
@typechecked
-def read_boolean(buf:bytes, bit_position: int) -> tuple[int, bool]:
+def read_boolean(buf:bytes|bytearray, bit_position: int) -> tuple[int, bool]:
    """
    Read a boolean value from a byte buffer.
@@ -60,7 +61,7 @@ def read_boolean(buf:bytes, bit_position: int) -> tuple[int, bool]:
    return bit_position, b==1
@typechecked
-def read_bits(buf:bytes, bit_position: int, nb_bits: int)  -> tuple[int, int]:
+def read_bits(buf:bytes|bytearray, bit_position: int, nb_bits: int)  -> tuple[int, int]:
    v = 0
    for _ in range(nb_bits):
        bit_position, bit = read_bit(buf, bit_position)
@@ -68,22 +69,22 @@ def read_bits(buf:bytes, bit_position: int, nb_bits: int)  -> tuple[int, int]:
    return bit_position, v
@typechecked
-def read_byte(buf:bytes, bit_position: int)  -> tuple[int, int]:
+def read_byte(buf:bytes|bytearray, bit_position: int)  -> tuple[int, int]:
    bit_position, b = read_bits(buf, bit_position, 8)
    return bit_position, b
@typechecked
-def read_word(buf:bytes, bit_position: int) -> tuple[int, int]:
+def read_word(buf:bytes|bytearray, bit_position: int) -> tuple[int, int]:
    bit_position, w = read_bits(buf, bit_position, 16)
    return bit_position, w
@typechecked
-def read_long(buf:bytes, bit_position: int) -> tuple[int, int]:
+def read_long(buf:bytes|bytearray, bit_position: int) -> tuple[int, int]:
    bit_position, value = read_bits(buf, bit_position, 32)
    return bit_position, value
@typechecked
-def read_unsigned_exp_golomb(buf:bytes, bit_position: int) -> tuple[int, int]:
+def read_unsigned_exp_golomb(buf:bytes|bytearray, bit_position: int) -> tuple[int, int]:
    nb_zeroes=0
    while True:
        bit_position, b = read_bit(buf, bit_position)
@@ -97,7 +98,7 @@ def read_unsigned_exp_golomb(buf:bytes, bit_position: int) -> tuple[int, int]:
    return bit_position, v-1
@typechecked
-def read_signed_exp_golomb(buf:bytes, bit_position: int) -> tuple[int, int]:
+def read_signed_exp_golomb(buf:bytes|bytearray, bit_position: int) -> tuple[int, int]:
    bit_position, v = read_unsigned_exp_golomb(buf, bit_position)
    match v%2:
        case 0:
@@ -172,7 +173,7 @@ def write_signed_exp_golomb(buf:bytearray, bit_position: int, v) -> int:
    return bit_position
@typechecked
-def parse_rbsp_trailing_bits(buf:bytes, bit_position: int) -> int:
+def parse_rbsp_trailing_bits(buf:bytes|bytearray, bit_position: int) -> int:
    bit_position, one = read_bit(buf, bit_position)
    if one==0:
        raise ValueError(f'Stop bit should be equal to one. Read: {one:d}')
@@ -192,7 +193,7 @@ def write_rbsp_trailing_bits(buf:bytearray, bit_position: int) -> int:
    return bit_position
@typechecked
-def more_rbsp_data(buf:bytes, bit_position: int) -> bool:
+def more_rbsp_data(buf:bytes|bytearray, bit_position: int) -> bool:
    logger.debug('Is there more data in buffer of length: %d at bit position: %d',
                 len(buf), bit_position)
@@ -249,7 +250,7 @@ def sodb_to_rbsp(buf:bytes) -> bytes:
 # Useful for SPS and PPS
@typechecked
-def parse_scaling_list(buf:bytes, bit_position: int, size) -> tuple[int,list[int]]:
+def parse_scaling_list(buf:bytes|bytearray, bit_position: int, size) -> tuple[int,list[int]]:
    res = []
    last_scale = 8
    next_scale = 8
@@ -284,7 +285,7 @@ def write_scaling_list(buf:bytearray, bit_position: int, size, matrix:list[int],
            bit_position = write_signed_exp_golomb(buf, bit_position, delta)
    else:
        logger.error('Not yet implemented')
-        exit(-1)
+        raise UnimplementedFeatureError("Optimized scaling list of H264 header is not implemented")
        # reverse = deltas.reverse()
        # compressed = False
        # while len(reverse)>0:
@@ -4,12 +4,14 @@
 import logging
 from dataclasses import dataclass, field
-from math import floor, log, ceil
+from math import ceil, floor, log
 from typeguard import typechecked
 import hexdump
 from tscut.h264.bitstream import (
    more_rbsp_data,
    parse_rbsp_trailing_bits,
    parse_scaling_list,
    read_bit,
    read_bits,
    read_boolean,
@@ -28,9 +30,6 @@ from tscut.h264.bitstream import (
    write_signed_exp_golomb,
    write_unsigned_exp_golomb,
    write_word,
    parse_scaling_list,
    parse_rbsp_trailing_bits,
    more_rbsp_data
 )
 logger = logging.getLogger(__name__)
@@ -6,20 +6,18 @@
 import logging
 from typing import IO
 from typeguard import typechecked
 import hexdump
 from typeguard import typechecked
 from tscut.exceptions import InvalidMediaError
 from tscut.h264.avc import AVCDecoderConfiguration, get_avc_config_from_h264
 from tscut.matroska.ebml import get_ebml_length
 from tscut.h264.avc import (
    AVCDecoderConfiguration,
    get_avc_config_from_h264
 )
 logger = logging.getLogger(__name__)
@typechecked
-def dump_codec_private_data(avc_decoder_configuration: AVCDecoderConfiguration) -> bytearray | None:
+def dump_codec_private_data(avc_decoder_configuration: AVCDecoderConfiguration) -> bytearray:
    # Rebuild a Matroska Codec Private Element
    res = bytearray()
    # Code private element
@@ -30,7 +28,7 @@ def dump_codec_private_data(avc_decoder_configuration: AVCDecoderConfiguration)
    embl_length = get_ebml_length(len(buf))
    if embl_length is None:
-        return None
+        raise InvalidMediaError
    logger.debug('EMBL encoded length: %s', hexdump.dump(embl_length, sep=':'))
    res.extend(embl_length)
    res.extend(buf)
@@ -38,7 +36,7 @@ def dump_codec_private_data(avc_decoder_configuration: AVCDecoderConfiguration)
    return res
@typechecked
-def get_codec_private_data_from_h264(input_file: IO[bytes]) -> bytearray | None:
+def get_codec_private_data_from_h264(input_file: IO[bytes]) -> bytearray:
    avcconfig = get_avc_config_from_h264(input_file)
    res = dump_codec_private_data(avcconfig)
@@ -13,8 +13,8 @@ from os import (
 )
 from typing import IO
 from typeguard import typechecked
 import hexdump
 from typeguard import typechecked
 logger = logging.getLogger(__name__)
@@ -0,0 +1,61 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 from dataclasses import dataclass
 from datetime import timedelta
 from enum import IntEnum, unique
 from pathlib import Path
 from typing import IO, BinaryIO
 from tscut.h264.avc import AVCDecoderConfiguration
@unique
 class SupportedFormat(IntEnum):
    TS = 1
    MP4 = 2
    MATROSKA = 3
    def __str__(self):
        match self:
            case SupportedFormat.TS:
                return 'mpegts'
            case SupportedFormat.MP4:
                return 'mov,mp4,m4a,3gp,3g2,mj2'
            case SupportedFormat.MATROSKA:
                return 'matroska,webm'
            case _:
                return 'Unsupported format'
@dataclass
 class ProcessingOptions:
    input_file: Path
    output_file: Path
    parts:  list[tuple[timedelta, timedelta]]
    tools_paths: dict[str, str]
    all_optional_tools: bool
    framerate: int|None = None
    threshold: int = 0
    coarse: bool = False
    subtitles_ocr: bool = True
    verbose: bool = False
    dump_memory: bool = False
    keep_temporaries: bool = False
@dataclass
 class PreparedMedia:
    basename: str
    movie: BinaryIO
    duration: timedelta
    framerate: float
    streams: list
    width: int
    height: int
    avc_config: AVCDecoderConfiguration
@dataclass
 class CutResult:
    filename: str
    movie: IO[bytes]
    check_positions: list[timedelta]
@@ -0,0 +1,253 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 import logging
 import os
 from datetime import datetime, timedelta
 from shutil import copyfile, move
 from typing import BinaryIO
 import hexdump
 from tscut.cutting import cut_recording
 from tscut.exceptions import InvalidMediaError, TemporaryFileError
 from tscut.h264.avc import parse_codec_private
 from tscut.matroska.codec import dump_codec_private_data
 from tscut.models import PreparedMedia, ProcessingOptions, SupportedFormat
 from tscut.subtitles.ocr import do_ocr, extract_srt, get_tesseract_supported_lang
 from tscut.temporaries import TemporaryFiles
 from tscut.tools.ffmpeg import ffmpeg_convert
 from tscut.tools.ffprobe import (
    find_subtitles_tracks,
    get_format,
    get_frame_rate,
    get_movie_duration,
    get_streams,
 )
 from tscut.tools.mkvtoolnix import (
    get_codec_private_data_from_mkv,
    remux_srt_subtitles,
 )
 logger = logging.getLogger(__name__)
 def detect_supported_format(format_name: str) -> SupportedFormat:
    for media_format in SupportedFormat:
        if format_name == str(media_format):
            return media_format
    raise InvalidMediaError(
        f"Unsupported media format: {format_name}"
    )
 def process_recording(options: ProcessingOptions, temporaries: TemporaryFiles) -> None:
    nb_parts = len(options.parts)
    basename = os.path.splitext(os.path.basename(options.input_file))[0]
    mp4filename = basename+'.mp4'
    mkvfilename = basename+'.mkv'
    try:
        input_file = open(options.input_file, mode='rb')
        logger.debug("Type of input file: %s", type(input_file))
    except OSError as e:
        raise InvalidMediaError(f"Impossible to open {options.input_file}") from e
    format_of_file = get_format(options.tools_paths['ffprobe'], input_file)
    if format_of_file is None:
        raise InvalidMediaError(f"Impossible to retrieve format of file: {input_file}")
    if 'format_name' not in format_of_file:
        raise InvalidMediaError(f"Impossible to retrieve format of file: {input_file}")
    if 'duration' not in format_of_file:
        raise InvalidMediaError(f"Impossible to retrieve duration of file: {input_file}")
    format_name = format_of_file['format_name']
    duration = timedelta(seconds=float(format_of_file['duration']))
    logger.info("Durée de l'enregistrement: %s", duration)
    if options.framerate is None:
        framerate = get_frame_rate(options.tools_paths['ffprobe'], input_file)
        if framerate is None:
            raise InvalidMediaError('Impossible to estimate frame rate !')
    else:
        framerate = options.framerate
    logger.info('Frame rate: %.1f fps', framerate)
    final_format_of_file = detect_supported_format(format_name)
    mkv: BinaryIO
    if final_format_of_file == SupportedFormat.TS:
        logger.info("Converting TS to MP4 (to fix timestamps).")
        try:
            with open(mp4filename, 'wb+') as mp4:
                ffmpeg_convert(options.tools_paths['ffmpeg'], options.tools_paths['ffprobe'],
                               input_file, 'mpegts', mp4, 'mp4', duration)
                temporaries.add(mp4)
                logger.info("Converting MP4 to MKV.")
                try:
                    mkv = open(mkvfilename, 'wb+')
                except OSError as e:
                    raise TemporaryFileError(f"Impossible to create: {mkvfilename}") from e
                ffmpeg_convert(options.tools_paths['ffmpeg'], options.tools_paths['ffprobe'],
                               mp4, 'mp4', mkv, 'matroska', duration)
                if nb_parts > 0:
                    temporaries.add(mkv)
        except OSError as e:
            raise TemporaryFileError(f"Impossible to create: {mp4filename}") from e
    elif final_format_of_file == SupportedFormat.MP4:
        logger.info("Converting MP4 to MKV")
        try:
            mkv = open(mkvfilename, 'wb+')
        except OSError as e:
            raise TemporaryFileError(f"Impossible to create: {mkvfilename}") from e
        ffmpeg_convert(options.tools_paths['ffmpeg'], options.tools_paths['ffprobe'],
                       input_file, 'mp4', mkv, 'matroska', duration)
        if nb_parts > 0:
            temporaries.add(mkv)
    else:
        logger.info("Already in MKV")
        mkv = input_file
    streams = get_streams(options.tools_paths['ffprobe'], mkv)
    if streams is None:
        raise InvalidMediaError(f"No streams found in file: {mkv}")
    logger.debug('Streams: %s', streams)
    main_video = None
    nb_videos = 0
    for stream in streams:
        if stream['codec_type'] == 'video':
            if stream['disposition']['default'] == 1:
                main_video = stream
                width = stream['width']
                height = stream['height']
                break
            nb_videos+=1
            if nb_videos == 1:
                main_video = stream
                width = stream['width']
                height = stream['height']
            else:
                main_video = None
    if main_video is None:
        raise InvalidMediaError("Impossible to find main video stream.")
    # We retrieve the main private codec data
    _, main_codec_private_data = get_codec_private_data_from_mkv(
                                                    mkvinfo_path=options.tools_paths['mkvinfo'],
                                                    input_file=mkv
                                                    )
    logger.debug('Main video stream has following private data: %s',
                 hexdump.dump(main_codec_private_data, sep=':'))
    # We parse them
    main_avc_config = parse_codec_private(main_codec_private_data)
    logger.debug('AVC configuration: %s', main_avc_config)
    # We check if the parse and dump operations are idempotent.
    private_data = dump_codec_private_data(main_avc_config)
    logger.debug('Redump AVC configuration: %s', hexdump.dump(private_data, sep=':'))
    # In rare occasion, the PPS has trailing zeroes that do not seem to be related to useful data
    # but they differ from the private data we generate that do not contain them.
    # In that case we try to redecode our own private data to see if both AVC configurations are
    # the same.
    if main_codec_private_data != private_data:
        logger.warning('Difference detected in bitstream !!')
        iso_avc_config = parse_codec_private(private_data)
        logger.debug('Reread AVC configuration: %s', iso_avc_config)
        # If there exists a difference between our own reconstructed AVC configuration and the
        # original one, we abandon
        if iso_avc_config != main_avc_config:
            raise InvalidMediaError(f"AVC configurations are different:\
                                        {main_avc_config}\n{iso_avc_config}\n")
    prepared_movie = PreparedMedia(
                        basename = basename,
                        movie = mkv,
                        duration = duration,
                        framerate = framerate,
                        streams = streams,
                        width = int(width),
                        height = int(height),
                        avc_config = main_avc_config
                    )
    cut_movie = cut_recording(prepared_movie, options, temporaries)
    final_with_video_name = cut_movie.filename
    final_with_video = cut_movie.movie
    checks = cut_movie.check_positions
    if options.subtitles_ocr:
        if not options.all_optional_tools:
            logger.warning("Missing tools for extracting subtitles.")
            move(final_with_video_name, options.output_file)
        else:
            # Final cut is not any more the final step.
            temporaries.add(final_with_video)
            duration = get_movie_duration(options.tools_paths['ffprobe'], final_with_video)
            supported_langs = get_tesseract_supported_lang(options.tools_paths['tesseract'])
            logger.info('Supported lang: %s', supported_langs)
            logger.info('Find subtitles tracks and language.')
            subtitles = find_subtitles_tracks(options.tools_paths['ffprobe'], final_with_video)
            logger.info(subtitles)
            sts: dict[str, list[int]] = {}
            for subtitle in subtitles:
                index = subtitle['index']
                if 'tags' in subtitle:
                    if 'language' in subtitle['tags']:
                        lang = subtitle['tags']['language']
                        if lang in sts:
                            sts[lang].append(index)
                        else:
                            sts[lang] = [index]
                    else:
                        logger.error("Dropping subtitle: %s because it is missing language\
                            indication", subtitle)
                else:
                    logger.error("Dropping subtitle: %s because it is missing language indication",
                                 subtitle)
            logger.info(sts)
            if len(sts) > 0:
                logger.info('Supported languages: %s', supported_langs)
                list_of_subtitles = extract_srt(options.tools_paths['mkvextract'],
                                                final_with_video_name, sts, supported_langs)
                logger.info(list_of_subtitles)
                for idx_name, sub_name, _, _ in list_of_subtitles:
                    try:
                        idx = open(idx_name,'rb')
                    except OSError as e:
                        raise TemporaryFileError(f"Impossible to open {idx_name}") from e
                    try:
                        sub = open(sub_name,'rb')
                    except OSError as e:
                        raise TemporaryFileError(f"Impossible to open {sub_name}") from e
                    temporaries.add(idx)
                    temporaries.add(sub)
                ocr = do_ocr(options.tools_paths['vobsubocr'], list_of_subtitles, duration,
                             temporaries, options.dump_memory)
                logger.info(ocr)
                # Remux SRT subtitles
                remux_srt_subtitles(options.tools_paths['mkvmerge'], final_with_video,
                                    options.output_file, ocr)
            else:
                copyfile(final_with_video_name, options.output_file)
    else:
        move(final_with_video_name, options.output_file)
    d = datetime(1,1,1)
    for c in checks:
        logger.info("Please check cut smoothness at %s", (c+d).strftime("%H:%M:%S"))
@@ -4,30 +4,33 @@
 import logging
-from typing import IO
+import os
-from io import TextIOWrapper
+import re
 from datetime import timedelta
 from io import TextIOWrapper
 from os import (
    read,
    SEEK_SET,
    fstat,
    lseek,
    memfd_create,
    read,
    write,
    fstat,
 )
 import re
 import os
 from subprocess import PIPE, Popen
 from typing import Sequence
 from iso639 import Lang
 from iso639.exceptions import InvalidLanguageValue
 from typeguard import typechecked
 from tqdm import tqdm
 from typeguard import typechecked
 from tscut.exceptions import ExternalToolError
 from tscut.temporaries import TemporaryFiles
 logger = logging.getLogger(__name__)
@typechecked
-def get_tesseract_supported_lang(tesseract_path:str) -> dict[Lang, str]|None:
+def get_tesseract_supported_lang(tesseract_path:str) -> dict[Lang, str]:
    """
    Retrieves the set of natural languages supported by the Tesseract OCR tool.
@@ -46,6 +49,7 @@ def get_tesseract_supported_lang(tesseract_path:str) -> dict[Lang, str]|None:
    res = {}
    with Popen([tesseract_path, '--list-langs'], stdout=PIPE) as tesseract:
        assert tesseract.stdout is not None
        for line in tesseract.stdout:
            line = line.decode('utf8')
            p = re.compile('(?P<lang>[a-z]{3})\n')
@@ -61,8 +65,7 @@ def get_tesseract_supported_lang(tesseract_path:str) -> dict[Lang, str]|None:
    tesseract.wait()
    if tesseract.returncode != 0:
-        logger.error("Tesseract returns an error code: %d",tesseract.returncode)
+        raise ExternalToolError("Tesseract returns an error code: %d",tesseract.returncode)
        return None
    return res
@@ -70,7 +73,7 @@ def get_tesseract_supported_lang(tesseract_path:str) -> dict[Lang, str]|None:
@typechecked
 def extract_srt(mkvextract:str, filename:str, subtitles:dict[str, list[int]],
-                langs:dict[Lang,str]) -> list[tuple[str,str,str,str]]|None:
+                langs:dict[Lang,str]) -> list[tuple[str,str,str,str]]:
    params = [mkvextract, filename, 'tracks']
    res = []
@@ -98,6 +101,7 @@ def extract_srt(mkvextract:str, filename:str, subtitles:dict[str, list[int]],
    env = {**os.environ, 'LANG': 'C'}
    with Popen(params, stdout=PIPE, close_fds=False, env=env) as extract:
        assert extract.stdout is not None
        pb = tqdm(TextIOWrapper(extract.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Extraction:')
        for line in pb:
@@ -106,7 +110,8 @@ def extract_srt(mkvextract:str, filename:str, subtitles:dict[str, list[int]],
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.update(int(m['progress'])-pb.n)
+                else:
                    pb.update(int(m['progress'])-pb.n)
        pb.update(100-pb.n)
        pb.refresh()
        pb.close()
@@ -120,14 +125,13 @@ def extract_srt(mkvextract:str, filename:str, subtitles:dict[str, list[int]],
        case 1:
            logger.warning('Mkvextract returns warning')
        case 2:
-            logger.error('Mkvextract returns an error code: %d', extract.returncode)
+            raise ExternalToolError('Mkvextract returns an error code: %d', extract.returncode)
            res = None
    return res
@typechecked
-def do_ocr(vobsubocr:str, idxs: list[tuple[str,str,str,str]], duration:timedelta,
+def do_ocr(vobsubocr:str, idxs: Sequence[tuple[str,str,str,str]], duration:timedelta,
-           temporaries:list[IO[bytes]], dump_mem_fd:bool=False):
+           temporaries:TemporaryFiles, dump_mem_fd:bool=False):
    res = []
    for idx_name, _, lang, iso in idxs:
@@ -140,6 +144,7 @@ def do_ocr(vobsubocr:str, idxs: list[tuple[str,str,str,str]], duration:timedelta
        srtfd = memfd_create(srtname, flags=0)
        with Popen([vobsubocr, '--lang', iso, idx_name], stdout=PIPE) as ocr:
            assert ocr.stdout is not None
            pb = tqdm(TextIOWrapper(ocr.stdout, encoding="utf-8"), total=
                      int(duration/timedelta(seconds=1)), unit='s', desc='OCR')
            for line in pb:
@@ -152,7 +157,7 @@ def do_ocr(vobsubocr:str, idxs: list[tuple[str,str,str,str]], duration:timedelta
                m = re.match(timestamps, line)
                if m is not None:
                    hours = int(m.group('hours'))
-                    minutes = int(m.group('hours'))
+                    minutes = int(m.group('minutes'))
                    seconds = int(m.group('seconds'))
                    ts = timedelta(hours=hours, minutes=minutes, seconds=seconds)
                    pb.n = int(ts/timedelta(seconds=1))
@@ -173,7 +178,7 @@ def do_ocr(vobsubocr:str, idxs: list[tuple[str,str,str,str]], duration:timedelta
                    pos = 0
                    while pos < srt_length:
                        pos+=write(outfd, buf[pos:])
-                        temporaries.append(dump_srt)
+                temporaries.add(dump_srt)
            except OSError:
                logger.error('Impossible to create file: %s', srtname)
                return None
@@ -0,0 +1,38 @@
 # SPDX-License-Identifier: GPL-2.0-or-later
 #
 # Copyright (C) 2026 Frédéric Tronel
 """A class to manage (and clean) all temporary files created during conversion"""
 import logging
 import os
 from typing import IO, Any, Self
 logger = logging.getLogger(__name__)
 class TemporaryFiles:
    def __init__(self, keep: bool = False) -> None:
        self._files: list[IO[Any]] = []
        self._keep = keep
    def add(self, file: IO[Any]) -> None:
        self._files.append(file)
    def cleanup(self) -> None:
        if self._keep:
            return
        for f in self._files:
            path = os.path.realpath(f.name)
            logger.info("Removing: %s", path)
            try:
                f.close()
                os.unlink(path)
            except OSError:
                logger.exception("Unable to remove temporary file: %s", path)
    def __enter__(self) -> Self:
        return self
    def __exit__(self) -> None:
        self.cleanup()
@@ -8,10 +8,10 @@ from shutil import which
 from typeguard import typechecked
 from tscut.exceptions import MissingToolError
 logger = logging.getLogger(__name__)
 class MissingToolError(Exception):
    pass
@typechecked
 def check_required_tools() -> tuple[bool,dict[str,str]]:
@@ -3,38 +3,32 @@
 # Copyright (C) 2026 Frédéric Tronel
 import logging
 from math import floor, ceil, log
 from datetime import timedelta
 from os import (
    read,
    SEEK_SET,
    lseek,
    memfd_create,
    set_inheritable,
    write,
    close
 )
 from typing import IO
 from subprocess import PIPE, Popen
 from io import TextIOWrapper
 from math import ceil, floor, log
 from os import SEEK_SET, close, lseek, memfd_create, read, set_inheritable, write
 from subprocess import PIPE, Popen
 from typing import IO, BinaryIO
 from typeguard import typechecked
 from tqdm import tqdm
 from typeguard import typechecked
 from tscut.exceptions import ExternalToolError, InvalidMediaError, TemporaryFileError
 from tscut.temporaries import TemporaryFiles
 from tscut.tools.ffprobe import (
    get_frames_in_stream,
    get_video_dimensions,
    with_subtitles,
    get_frames_in_stream,
 )
 from tscut.tools.ppm import dump_ppm
-from tscut.tools.timeframe import parse_timestamp, get_packet_duration
+from tscut.tools.timeframe import get_packet_duration, parse_timestamp
 logger = logging.getLogger(__name__)
@typechecked
 def ffmpeg_convert(ffmpeg_path:str, ffprobe_path:str, input_file: IO[bytes], input_format:str,
-                   output_file: IO[bytes], output_format:str, duration: timedelta):
+                   output_file: IO[bytes], output_format:str, duration: timedelta) -> None:
    width, height =  get_video_dimensions(ffprobe_path, input_file)
    if width is None or height is None:
        return
@@ -46,10 +40,7 @@ def ffmpeg_convert(ffmpeg_path:str, ffprobe_path:str, input_file: IO[bytes], inp
    set_inheritable(infd, True)
    set_inheritable(outfd, True)
-    if logger.getEffectiveLevel() == logging.DEBUG:
+    log_level = [] if logger.getEffectiveLevel() == logging.DEBUG else ['-loglevel', 'quiet']
        log_level = []
    else:
        log_level = [ '-loglevel', 'quiet' ]
    params = [ffmpeg_path, '-y',]+log_level+['-progress', '/dev/stdout', '-canvas_size',
                                  f'{width:d}x{height:d}', '-f', input_format,
@@ -66,6 +57,7 @@ def ffmpeg_convert(ffmpeg_path:str, ffprobe_path:str, input_file: IO[bytes], inp
    logger.debug('Executing %s',  params)
    with Popen(params, stdout=PIPE, close_fds=False) as ffmpeg:
        assert ffmpeg.stdout is not None
        pb = tqdm(TextIOWrapper(ffmpeg.stdout, encoding="utf-8"),
                  total=int(duration/timedelta(seconds=1)), unit='s', desc='Conversion')
        for line in pb:
@@ -77,12 +69,12 @@ def ffmpeg_convert(ffmpeg_path:str, ffprobe_path:str, input_file: IO[bytes], inp
                    pb.update()
        status = ffmpeg.wait()
        if status != 0:
-            logger.error('Conversion failed with status code: %d', status)
+            raise ExternalToolError(f"Conversion failed with status code: {status:d}")
@typechecked
 def extract_pictures(ffmpeg_path:str, input_file:IO[bytes], begin:timedelta, nb_frames:int,
-                     width:int=640, height:int=480) -> tuple[bytes,int]|tuple[None,None]:
+                     width:int=640, height:int=480) -> tuple[bytes,int]:
    """
    Extract pictures from a video file using FFmpeg.
@@ -125,14 +117,12 @@ def extract_pictures(ffmpeg_path:str, input_file:IO[bytes], begin:timedelta, nb_
    with Popen(command, stdout=PIPE, close_fds=False) as ffmpeg:
        status = ffmpeg.wait()
        if status != 0:
-            logger.error('Conversion failed with status code: %d', status)
+            raise ExternalToolError(f"Conversion failed with status code: {status:d}")
            return None, None
        lseek(outfd, 0, SEEK_SET)
        images = read(outfd,length)
        if len(images) != length:
-            logger.error("Received %d bytes but %d were expected.", len(images), length)
+            raise InvalidMediaError(f"Received {len(images)} bytes but {length} were expected.")
            return None, None
    lseek(outfd, 0, SEEK_SET)
    return images, outfd
@@ -141,7 +131,7 @@ def extract_pictures(ffmpeg_path:str, input_file:IO[bytes], begin:timedelta, nb_
 def extract_sound(ffmpeg_path:str, input_file: IO[bytes], begin:timedelta, output_filename:str,
                  packet_duration:int, sub_channel:int=0,
                  nb_packets:int=0, sample_rate:int=48000, 
-                  nb_channels:int=2) -> tuple[bytes,int]|tuple[None,None]:
+                  nb_channels:int=2) -> tuple[bytes,int]:
    outfd = memfd_create(output_filename, flags=0)
    infd = input_file.fileno()
    lseek(infd, 0, SEEK_SET)
@@ -159,24 +149,25 @@ def extract_sound(ffmpeg_path:str, input_file: IO[bytes], begin:timedelta, outpu
    with Popen(command, stdout=PIPE, close_fds=False) as ffmpeg:
        status = ffmpeg.wait()
        if status != 0:
-            logger.error('Sound extraction returns error code: %d', status)
+            raise ExternalToolError(f"Sound extraction returns error code: {status}")
            return None, None
        lseek(outfd, 0, SEEK_SET)
        sound = read(outfd, length)
        if len(sound) != length:
-            logger.info("Received %d bytes but %d were expected (channels=%d, freq=%d, packets=%d,\
+            raise InvalidMediaError(f"Received {len(sound)} bytes but {length} were expected (\
-                        duration=%d ms).", len(sound), length, nb_channels, sample_rate, nb_packets,
+                            channels={nb_channels}, freq={sample_rate} packets={nb_packets},\
-                        packet_duration)
+                        duration={packet_duration} ms).")
            return None, None
        return sound, outfd
@typechecked
 def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes], begin:timedelta,
                        end:timedelta, streams, files_prefix, nb_frames:int, framerate:float,
-                        width:int, height:int, temporaries, dump_mem_fd:bool=False):
+                        width:int, height:int, temporaries:TemporaryFiles,
                        dump_mem_fd:bool=False) -> tuple[BinaryIO|None,
                                                         TextIOWrapper|None,
                                                         BinaryIO|None]:
    # The command line for encoding only video track
    video_encoder_params = [ ffmpeg_path, '-y', '-loglevel', 'quiet']
    video_input_params = []
@@ -236,9 +227,6 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
                images_bytes, memfd = extract_pictures(ffmpeg_path, input_file=input_file,
                                                       begin=begin, nb_frames=nb_frames,
                                                       width=width, height=height)
                if images_bytes is None or memfd is None:
                    logger.error('Impossible to extract picture from video stream.')
                    exit(-1)
                memfds.append(memfd)
                if dump_mem_fd:
@@ -253,7 +241,7 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
                video_codec_params.extend([f'-c:v:{video_id:d}', codec, f'-level:v:{video_id:d}',
                                           level, '-pix_fmt', pixel_format])
                video_codec_params.extend(interlaced_options)
-                video_codec_params.extend([f'-colorspace:v:{video_id}', color_space,
+                video_codec_params.extend([f'-colorspace:v:{video_id:d}', color_space,
                                         f'-color_primaries:v:{video_id:d}', color_primaries,
                                         f'-color_trc:v:{video_id:d}', color_transfer,
                                         f'-color_range:v:{video_id:d}', color_range])
@@ -262,26 +250,19 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
                logger.debug('Audio stream: %s', stream)
                sample_rate = int(stream['sample_rate'])
                nb_channels = int(stream['channels'])
-                if 'bit_rate' in stream:
+                bit_rate = int(stream['bit_rate']) if 'bit_rate' in stream else 128000
                    bit_rate = int(stream['bit_rate'])
                else:
                    bit_rate = 128000
                codec = stream['codec_name']
-                if 'tags' in stream:
+                if 'tags' in stream and 'language' in stream['tags']:
-                    if 'language' in stream['tags']:
+                    generic_codec_params.extend([f'-metadata:s:a:{audio_id:d}',
                        generic_codec_params.extend([f'-metadata:s:a:{audio_id:d}',
                                                     f"language={stream['tags']['language']}"])
                packets = get_frames_in_stream(ffprobe_path, input_file=input_file, begin=begin,
                                               end=end, stream_kind='a', sub_stream_id=audio_id)
                if packets is None:
-                    logger.error("Impossible to retrieve audio packets")
+                    raise InvalidMediaError("Impossible to retrieve audio packets")
                    return None
                nb_packets = len(packets)
                logger.debug("Found %d packets to be extracted from audio track.", nb_packets)
                if nb_packets > 0:
                    packet_duration = get_packet_duration(packets[0])
                    if packet_duration is None:
                        return None
                else:
                    packet_duration = 0
@@ -294,23 +275,18 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
                                                   output_filename=tmpname,
                                                   sample_rate=sample_rate, nb_channels=nb_channels)
                if sound_bytes is None or memfd is None:
                    logger.error('Impossible to extract sound track')
                    exit(-1)
                memfds.append(memfd)
                if dump_mem_fd:
                    try:
                        with open(tmpname,'wb') as output:
-                            temporaries.append(output)
+                            temporaries.add(output)
                            outfd = output.fileno()
                            pos = 0
                            while pos < len(sound_bytes):
                                pos+=write(outfd, sound_bytes[pos:])
-                    except OSError:
+                    except OSError as e:
-                        logger.error('Impossible to create file: %s', tmpname)
+                        raise TemporaryFileError(f"Impossible to create file: {tmpname}") from e
                        return None
                # We rewind to zero the memory file descriptor
                lseek(memfd, 0, SEEK_SET)
@@ -325,9 +301,8 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
                logger.info("Extracting a subtitle stream: s:%d", subtitle_id)
                codec = stream['codec_name']
                generic_input_params.extend(['-i', './empty.idx'])
-                if 'tags' in stream:
+                if 'tags' in stream and 'language' in stream['tags']:
-                    if 'language' in stream['tags']:
+                    generic_codec_params.extend([f'-metadata:s:s:{subtitle_id:d}',
                        generic_codec_params.extend([f'-metadata:s:s:{subtitle_id:d}',
                                                     f"language={stream['tags']['language']}"])
                generic_codec_params.extend([f'-c:s:{subtitle_id:d}', 'copy'])
                subtitle_id=subtitle_id+1
@@ -344,9 +319,8 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
        mkv_filename = f'{files_prefix}.mkv'
        try:
            mkv_output = open(mkv_filename,'wb+')
-        except OSError:
+        except OSError as e:
-            logger.error('Impossible to create file: %s', mkv_filename)
+            raise TemporaryFileError(f"Impossible to create file: {mkv_filename}") from e
            return None
        mkvoutfd = mkv_output.fileno()
        set_inheritable(mkvoutfd, True)
@@ -357,17 +331,15 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
        with Popen(generic_encoder_params, stdout=PIPE, close_fds=False) as ffmpeg:
            status = ffmpeg.wait()
            if status != 0:
-                logger.error('Encoding failed with status code: %d', status)
+                raise ExternalToolError(f"Encoding failed with status code: {status}")
                return None
-            temporaries.append(mkv_output)
+            temporaries.add(mkv_output)
        h264_filename = f'{files_prefix}.h264'
        try:
            h264_output = open(h264_filename,'wb+')
-        except OSError:
+        except OSError as e:
-            logger.error('Impossible to create file: %s', h264_filename)
+            raise TemporaryFileError(f"Impossible to create file {h264_filename}") from e
            return None
        h264outfd = h264_output.fileno()
        set_inheritable(h264outfd, True)
@@ -386,17 +358,15 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
        with Popen(video_encoder_params, stdout=PIPE, close_fds=False) as ffmpeg:
            status = ffmpeg.wait()
            if status != 0:
-                logger.error('Encoding failed with status code: %d', status)
+                raise ExternalToolError(f"Encoding failed with status code: {status:d}")
                return None
-            temporaries.append(h264_output)
+            temporaries.add(h264_output)
        h264_ts_filename = f'{files_prefix}-ts.txt'
        try:
            h264_ts_output = open(h264_ts_filename,'w+', encoding='utf8')
-        except OSError:
+        except OSError as e:
-            logger.error('Impossible to create file: %s', h264_ts_filename)
+            raise TemporaryFileError(f"Impossible to create file: {h264_ts_filename}") from e
            return None
        h264_ts_output.write('# timestamp format v2\n')
        ts = 0
@@ -406,7 +376,7 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
        h264_ts_output.flush()
        h264_ts_output.seek(0)
-        temporaries.append(h264_ts_output)
+        temporaries.add(h264_ts_output)
        for memfd in memfds:
            close(memfd)
@@ -414,5 +384,5 @@ def extract_all_streams(ffmpeg_path:str, ffprobe_path:str, input_file:IO[bytes],
        return h264_output, h264_ts_output, mkv_output
    # Nothing to be done. We are already at a i-frame boundary.
-    return None, None
+    return None, None, None
@@ -4,20 +4,21 @@
 import json
 import logging
 from math import floor
 from datetime import timedelta
 import os
 from datetime import timedelta
 from io import BytesIO
 from math import floor
 from os import (
    SEEK_SET,
    lseek,
    set_inheritable,
 )
 from typing import IO
 from subprocess import PIPE, Popen
-from io import BytesIO
+from typing import IO
 from typeguard import typechecked
 from tscut.exceptions import InvalidMediaError, ExternalToolError
 from tscut.tools.timeframe import get_ts_frame
 logger = logging.getLogger(__name__)
@@ -63,9 +64,8 @@ def get_frame_rate(ffprobe_path:str, input_file: IO[bytes]) -> float|None:
        out = json.load(BytesIO(out))
        if 'frames' in out:
            for frame in out['frames']:
-                if 'interlaced_frame' in frame:
+                if 'interlaced_frame' in frame and frame['interlaced_frame'] == 1:
-                    if frame['interlaced_frame'] == 1:
+                    interlaced = True
                        interlaced = True
                if 'pts_time' in frame:
                    ts = float(frame['pts_time'])
                    if min_ts is None:
@@ -155,7 +155,7 @@ def get_format(ffprobe_path:str, input_file: IO[bytes]) -> dict|None:
    return None
@typechecked
-def get_movie_duration(ffprobe_path:str, input_file: IO[bytes]) -> timedelta|None:
+def get_movie_duration(ffprobe_path:str, input_file: IO[bytes]) -> timedelta:
    infd = input_file.fileno()
    lseek(infd, 0, SEEK_SET)
    set_inheritable(infd, True)
@@ -165,16 +165,12 @@ def get_movie_duration(ffprobe_path:str, input_file: IO[bytes]) -> timedelta|Non
        out = json.load(BytesIO(out))
        if 'format' in out and 'duration' in out['format']:
            duration = floor(float(out['format']['duration']))
-            ts = timedelta(seconds=duration)
+            return timedelta(seconds=duration)
-            return ts
+        raise InvalidMediaError("Impossible to retrieve duration of movie")
        logger.error('Impossible to retrieve duration of movie')
    return None
 # ffprobe -loglevel quiet -select_streams v:0 -show_entries stream=width,height -of json sample.ts
@typechecked
-def get_video_dimensions(ffprobe_path:str,
+def get_video_dimensions(ffprobe_path:str, input_file: IO[bytes]) -> tuple[int,int]:
                         input_file: IO[bytes]) -> tuple[int,int]| tuple[None,None]:
    infd = input_file.fileno()
    lseek(infd, 0, SEEK_SET)
    set_inheritable(infd, True)
@@ -188,8 +184,7 @@ def get_video_dimensions(ffprobe_path:str,
            if ('width' in video) and ('height' in video):
                return int(video['width']), int(video['height'])
-    logger.error('Impossible to retrieve dimensions of video')
+    raise InvalidMediaError('Impossible to retrieve dimensions of video')
    return None, None
@typechecked
 def get_streams(ffprobe_path:str, input_file: IO[bytes]) -> list|None:
@@ -244,7 +239,7 @@ def with_subtitles(ffprobe_path:str, input_file: IO[bytes]) -> bool:
    return False
-def find_subtitles_tracks(ffprobe_path:str, input_file: IO[bytes]) -> dict|None:
+def find_subtitles_tracks(ffprobe_path:str, input_file: IO[bytes]) -> dict:
    infd = input_file.fileno()
    lseek(infd, 0, SEEK_SET)
    set_inheritable(infd, True)
@@ -259,10 +254,9 @@ def find_subtitles_tracks(ffprobe_path:str, input_file: IO[bytes]) -> dict|None:
        out = json.load(BytesIO(out))
        if 'streams' in out:
            return out['streams']
        logger.error('Impossible to retrieve format of file')
    ffprobe.wait()
-    return None
+    raise InvalidMediaError('Impossible to retrieve format of file')
@typechecked
 def get_frames_in_stream(ffprobe_path: str, input_file: IO[bytes], begin:timedelta, end:timedelta,
@@ -280,8 +274,7 @@ def get_frames_in_stream(ffprobe_path: str, input_file: IO[bytes], begin:timedel
        frames = json.load(BytesIO(out))
        status = ffprobe.wait()
        if status != 0:
-            logger.error('ffprobe failed with status code: %d', status)
+            raise ExternalToolError(f"ffprobe failed with status code: {status:d}")
            return None
        # Sort frames by timestamp
        tmp = {}
@@ -289,8 +282,6 @@ def get_frames_in_stream(ffprobe_path: str, input_file: IO[bytes], begin:timedel
            frames = frames['frames']
            for frame in frames:
                ts = get_ts_frame(frame)
                if ts is None:
                    return None
                if begin <= ts <= end:
                    tmp[ts]=frame
@@ -326,28 +317,25 @@ def get_nearest_idr_frame(ffprobe_path: str, input_file: IO[bytes], timestamp:ti
        frames = json.load(BytesIO(out))
        status = ffprobe.wait()
        if status != 0:
-            logger.error('ffprobe failed with status code: %d', status)
+            raise ExternalToolError(f"ffprobe failed with status code: {status:d}")
            return None
        if 'frames' in frames:
            frames = frames['frames']
            for frame in frames:
                ts = get_ts_frame(frame)
                if ts is None:
                    return None
                if tbegin <= ts <= tend:
                    idrs.append(frame)
        else:
            logger.error('Impossible to retrieve IDR frames inside file around [%s,%s]',
                         tbegin, tend)
-            return None
+            return
-    return None
+    return
@typechecked
 def get_nearest_iframe(ffprobe_path:str, input_file: IO[bytes],
                       timestamp:timedelta, before:bool=True,
-                       delta_max:timedelta=timedelta(seconds=15))-> tuple[int,dict | None]:
+                       delta_max:timedelta=timedelta(seconds=15))-> tuple[int,dict] | None:
    infd = input_file.fileno()
    set_inheritable(infd, True)
@@ -356,14 +344,8 @@ def get_nearest_iframe(ffprobe_path:str, input_file: IO[bytes],
    iframe = None
    while delta < delta_max:
        zero = timedelta()
-        if before:
+        tbegin = timestamp - delta if before else timestamp
-            tbegin = timestamp-delta
+        tend = timestamp + delta if not before else timestamp
        else:
            tbegin = timestamp
        if not before:
            tend = timestamp+delta
        else:
            tend = timestamp
        tbegin = max(tbegin, zero)
        logger.debug('Looking for an iframe in [%s, %s]', tbegin, tend)
@@ -381,8 +363,9 @@ def get_nearest_iframe(ffprobe_path:str, input_file: IO[bytes],
        found = False
        for frame in iframes:
-            ts = get_ts_frame(frame)
+            try:
-            if ts is None:
+                ts = get_ts_frame(frame)
            except InvalidMediaError:
                logger.warning('I-frame with no timestamp: %s', frame)
                continue
@@ -404,11 +387,12 @@ def get_nearest_iframe(ffprobe_path:str, input_file: IO[bytes],
        its = get_ts_frame(iframe)
        if its is None:
            logger.error("Impossible to retrieve timestamp of i-frame !")
-            return 0, None
+            return None
        nb_frames = 0
        for frame in frames:
-            ts = get_ts_frame(frame)
+            try:
-            if ts is None:
+                ts = get_ts_frame(frame)
            except InvalidMediaError:
                logger.warning('Frame without timestamp: %s', frame)
                continue
@@ -422,6 +406,6 @@ def get_nearest_iframe(ffprobe_path:str, input_file: IO[bytes],
                    nb_frames = nb_frames+1
    else:
        logger.error("Impossible to find I-frame between: %s and %s", tbegin, tend)
-        return 0, None
+        return None
    return(nb_frames, iframe)
@@ -1,125 +1,30 @@
-#!/usr/bin/env python3
+# SPDX-License-Identifier: GPL-2.0-or-later
-'''A module to remove parts of video (.e.g advertisements) with single frame precision.'''
+#
 # Copyright (C) 2026 Frédéric Tronel
 # Standard modules
 import json
 import logging
-import os.path
+import os
 import re
 from datetime import timedelta
-from enum import IntEnum, unique
+from io import TextIOWrapper
-from io import BytesIO, TextIOWrapper
+from os import SEEK_SET, fstat, ftruncate, lseek, read, set_inheritable, write
-from math import ceil, floor, log
+from pathlib import Path
 from os import (
    SEEK_SET,
    close,
    fstat,
    ftruncate,
    lseek,
    memfd_create,
    read,
    set_inheritable,
    write,
 )
 from subprocess import PIPE, Popen
-from sys import exit
+from typing import IO, Sequence
 from typing import IO
 # Third party libraries
 import hexdump
 from iso639 import Lang
 from iso639.exceptions import InvalidLanguageValue
 from tqdm import tqdm
 from typeguard import typechecked
-from tscut.h264.avc import (
+from tscut.exceptions import InvalidMediaError, ExternalToolError
    get_avc_config_from_h264,
    parse_codec_private
 )
 from tscut.matroska.ebml import change_ebml_element_size
 from tscut.matroska.codec import dump_codec_private_data
 from tscut.tools.ffprobe import (
    get_format,
    get_frame_rate,
    get_video_dimensions,
    get_movie_duration,
    get_streams,
    with_subtitles,
    find_subtitles_tracks,
    get_nearest_iframe,
    get_ts_frame
 )
 from tscut.tools.ffmpeg import (
    extract_all_streams,
    ffmpeg_convert
 )
 from tscut.tools.timeframe import (
    compare_time_interval,
    parse_time_interval
 )
 from tscut.tools.discovery import check_required_tools
 from tscut.ocr.subtitles import (
    do_ocr,
    extract_srt,
    get_tesseract_supported_lang
 )
 # Useful SPS/PPS discussion.
 # https://copyprogramming.com/howto/including-sps-and-pps-in-a-raw-h264-track
 # https://gitlab.com/mbunkus/mkvtoolnix/-/issues/2390
 # New strategy: a possible way of handling multiple SPS/PPS gracefully.
 # Encode each head and trailer with FFMPEG using only I-frame (to be sure the NAL unit will never
 # refer to another image).
 # Encode using an different SPS-ID all of them (using sps-id parameter of libx264 library, e.g
 # 1 instead of 0).
 # For the video track produce only a raw H264 file and a file containing timestamps of the
 # different frames.
 # For the rest of the tracks (audio, subtitles) produce directly a MKV (this is already done).
 # Concatenate all raw H264 in a giant one (like cat), and the same for timestamps of video frames
 # (to keep sound and video synchronized).
 # Then use mkvmerge to remux the H264 track and the rest of tracks.
 # MKVmerge "concatenate" subcommand is able to concatenate different SPS/PPS data into a bigger
 # Private Codec Data.
 # However, this is proved to be not reliable. Sometimes it results in a AVC context containing
 # a single SPS/PPS.
 # So we have to rely on a manual parsing of the H264 AVC context of original movie
 # and the ones produced for headers and trailers, and then merging them into a bigger AVC context.
 # Then finally, change the Private Codec Data in the final MKV.
@unique
 class SupportedFormat(IntEnum):
    TS = 1
    MP4 = 2
    MATROSKA = 3
    def __str__(self):
        match self:
            case SupportedFormat.TS:
                return 'mpegts'
            case SupportedFormat.MP4:
                return 'mov,mp4,m4a,3gp,3g2,mj2'
            case SupportedFormat.MATROSKA:
                return 'matroska,webm'
            case _:
                return 'Unsupported format'
 # Extract SPS/PPS
 # https://gitlab.com/mbunkus/mkvtoolnix/-/issues/2390
 # ffmpeg -i <InputFile (before concatenation)> -c:v copy -an -sn -bsf:v trace_headers -t 0.01\
 #        -report -loglevel 0 -f null -
 # Found codec private data using mkvinfo
@typechecked
-def get_codec_private_data_from_mkv(mkvinfo_path:str, 
+def get_codec_private_data_from_mkv(mkvinfo_path:str,
-                                    input_file: IO[bytes]) -> tuple[int, bytes]|tuple[None,None]:
+                                    input_file: IO[bytes]) -> tuple[int, bytes]:
    logger = logging.getLogger(__name__)
    infd = input_file.fileno()
@@ -135,11 +40,11 @@ def get_codec_private_data_from_mkv(mkvinfo_path:str,
    with Popen([mkvinfo_path, '-z', '-X', '-P', f'/proc/self/fd/{infd:d}'], stdout=PIPE,
               close_fds=False, env=env) as mkvinfo:
        out, _ = mkvinfo.communicate()
-        out = out.decode('utf8')
+        lines = out.decode('utf8')
        reg_exp = (r"^.*Codec's private data: size ([0-9]+) \(H.264.*\) hexdump "
                  r"(?P<hexdump>([0-9a-f]{2} )+)at (?P<position>[0-9]+) size (?P<size>[0-9]+).*$")
        p = re.compile(reg_exp)
-        for line in out.splitlines():
+        for line in lines.splitlines():
            m = p.match(line)
            if m is not None:
                size = int(m.group('size'))
@@ -154,7 +59,7 @@ def get_codec_private_data_from_mkv(mkvinfo_path:str,
        data = read(infd, size)
        return position, data
-    return None, None
+    raise InvalidMediaError("Impossible to retrieve codec private data from MKV !")
@typechecked
 def parse_mkv_tree(mkvinfo_path:str, input_file: IO[bytes]) -> dict[str,tuple[int,int]]:
@@ -169,13 +74,13 @@ def parse_mkv_tree(mkvinfo_path:str, input_file: IO[bytes]) -> dict[str,tuple[in
    with Popen([mkvinfo_path, '-z', '-X', '-P', f'/proc/self/fd/{infd:d}'], stdout=PIPE,
               close_fds=False, env=env) as mkvinfo:
        out, _ = mkvinfo.communicate()
-        out = out.decode('utf8')
+        lines = out.decode('utf8')
        prefix = []
        reg_exp = (r"(^(?P<root>\+)|(\|(?P<depth>[ ]*\+))).*at (?P<position>[0-9]+)"
                  r" size (?P<size>[0-9]+).*$")
        p = re.compile(reg_exp)
        prev_depth = -1
-        for line in out.splitlines():
+        for line in lines.splitlines():
            m = p.match(line)
            if m is None:
                logger.error("Impossible to match line: %s", line)
@@ -211,10 +116,9 @@ def parse_mkv_tree(mkvinfo_path:str, input_file: IO[bytes]) -> dict[str,tuple[in
    mkvinfo.wait()
    return elements
@typechecked
-def change_codec_private_data(mkvinfo_path:str, input_file: IO[bytes], codec_data:bytes) -> None:
+def change_codec_private_data(mkvinfo_path:str, input_file: IO[bytes],
                              codec_data:bytes | bytearray) -> None:
    logger = logging.getLogger(__name__)
    infd = input_file.fileno()
@@ -223,6 +127,8 @@ def change_codec_private_data(mkvinfo_path:str, input_file: IO[bytes], codec_dat
    current_length = fstat(infd).st_size
    logger.info('Current size of file: %d', current_length)
    position, current_data = get_codec_private_data_from_mkv(mkvinfo_path, input_file)
    if position is None or current_data is None:
        raise InvalidMediaError("Impossible to retrieve private data from file")
    current_data_length = len(current_data)
    future_length = current_length - current_data_length + len(codec_data)
    logger.info('Expected size of file: %d', future_length)
@@ -240,8 +146,7 @@ def change_codec_private_data(mkvinfo_path:str, input_file: IO[bytes], codec_dat
            break
    if not found:
-        logger.error('Impossible to retrieve the key of codec private data')
+        raise InvalidMediaError("Impossible to retrieve the key of codec private data")
        exit(-1)
    if current_length < future_length:
        lseek(infd, position+current_data_length, SEEK_SET)
@@ -284,10 +189,6 @@ def change_codec_private_data(mkvinfo_path:str, input_file: IO[bytes], codec_dat
            # been resized).
            delta+=change_ebml_element_size(input_file, pos, delta)
@typechecked
 def extract_mkv_part(mkvmerge_path:str, input_file:IO[bytes], output_file:IO[bytes],
                     begin:timedelta, end:timedelta) -> None:
@@ -308,6 +209,7 @@ def extract_mkv_part(mkvmerge_path:str, input_file:IO[bytes], output_file:IO[byt
    logger.debug('Executing: %s', command)
    with Popen(command, stdout=PIPE, close_fds=False, env=env) as mkvmerge:
        assert mkvmerge.stdout is not None
        pb = tqdm(TextIOWrapper(mkvmerge.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Extraction')
        for line in pb:
@@ -316,7 +218,8 @@ def extract_mkv_part(mkvmerge_path:str, input_file:IO[bytes], output_file:IO[byt
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.update(int(m['progress'])-pb.n)
+                else:
                    pb.update(int(m['progress'])-pb.n)
            elif line.startswith('Warning'):
                warnings.append(line)
        pb.update(100-pb.n)
@@ -329,14 +232,14 @@ def extract_mkv_part(mkvmerge_path:str, input_file:IO[bytes], output_file:IO[byt
            for w in warnings:
                logger.warning(w)
        elif status == 2:
-            logger.error('Extraction returns errors')
+            raise ExternalToolError("Extraction returns errors")
 # Merge a list of mkv files passed as input, and produce a new MKV as output
@typechecked
-def merge_mkvs(mkvmerge_path:str, inputs: list[IO[bytes]], output_name:str,
+def merge_mkvs(mkvmerge_path:str, inputs: Sequence[IO[bytes]], output_name:str,
-               concatenate: bool=True, timestamps: dict[int, IO[str]] | None = None) -> IO[bytes]|None:
+               concatenate: bool=True,
               timestamps: dict[int, IO[str]] | None = None) -> IO[bytes]:
    logger = logging.getLogger(__name__)
    if timestamps is None:
@@ -345,9 +248,9 @@ def merge_mkvs(mkvmerge_path:str, inputs: list[IO[bytes]], output_name:str,
    fds = []
    try:
        out = open(output_name, 'wb+')
-    except OSError:
+    except OSError as e:
        logger.error('Impossible to create file: %s', output_name)
-        return None
+        raise e
    outfd = out.fileno()
    lseek(outfd, 0, SEEK_SET)
@@ -390,6 +293,7 @@ def merge_mkvs(mkvmerge_path:str, inputs: list[IO[bytes]], output_name:str,
    logger.debug('Executing: LANG=C %s', merge_params)
    with Popen(merge_params, stdout=PIPE, close_fds=False, env=env) as mkvmerge:
        assert mkvmerge.stdout is not None
        pb = tqdm(TextIOWrapper(mkvmerge.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Merging')
        for line in pb:
@@ -398,8 +302,9 @@ def merge_mkvs(mkvmerge_path:str, inputs: list[IO[bytes]], output_name:str,
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.n = int(m['progress'])
+                else:
-                pb.update()
+                    pb.n = int(m['progress'])
                    pb.update()
            elif line.startswith('Warning'):
                warnings.append(line)
@@ -443,6 +348,7 @@ def extract_track_from_mkv(mkvextract_path: str, input_file: IO[bytes], index,
    logger.debug('Executing: LANG=C %s', params)
    with Popen(params, stdout=PIPE, close_fds=False, env=env) as extract:
        assert extract.stdout is not None
        pb = tqdm(TextIOWrapper(extract.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Extraction of track')
        for line in pb:
@@ -451,7 +357,8 @@ def extract_track_from_mkv(mkvextract_path: str, input_file: IO[bytes], index,
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.update(int(m['progress'])-pb.n)
+                else:
                    pb.update(int(m['progress'])-pb.n)
        pb.update(100-pb.n)
        pb.refresh()
        pb.close()
@@ -459,10 +366,12 @@ def extract_track_from_mkv(mkvextract_path: str, input_file: IO[bytes], index,
    extract.wait()
    if extract.returncode != 0:
-        logger.error('Mkvextract returns an error code: %d', extract.returncode)
+        raise ExternalToolError(f"Mkvextract returns an error code: {extract.returncode:d}")
    else:
        logger.info('Track %d was succesfully extracted.', index)
@typechecked
 def remove_video_tracks_from_mkv(mkvmerge_path:str, input_file: IO[bytes],
                                 output_file: IO[bytes]) -> None:
@@ -480,6 +389,7 @@ def remove_video_tracks_from_mkv(mkvmerge_path:str, input_file: IO[bytes],
    env = {**os.environ, 'LANG': 'C'}
    with Popen(params, stdout=PIPE, close_fds=False, env=env) as remove:
        assert remove.stdout is not None
        pb = tqdm(TextIOWrapper(remove.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Removal of video track:')
        for line in pb:
@@ -488,7 +398,8 @@ def remove_video_tracks_from_mkv(mkvmerge_path:str, input_file: IO[bytes],
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.update(int(m['progress'])-pb.n)
+                else:
                    pb.update(int(m['progress'])-pb.n)
        pb.update(100-pb.n)
        pb.refresh()
        pb.close()
@@ -496,19 +407,20 @@ def remove_video_tracks_from_mkv(mkvmerge_path:str, input_file: IO[bytes],
    remove.wait()
    if remove.returncode != 0:
-        logger.error('Mkvmerge returns an error code: %d', remove.returncode)
+        raise ExternalToolError(f"Mkvmerge returns an error code: {remove.returncode:d}")
    else:
        logger.info('Video tracks were succesfully extracted.')
@typechecked
-def remux_srt_subtitles(mkvmerge_path:str, input_file: IO[bytes], output_filename: str,
+def remux_srt_subtitles(mkvmerge_path:str, input_file: IO[bytes], output_file: Path,
                        subtitles) -> None:
    logger = logging.getLogger(__name__)
    try:
-        out = open(output_filename, 'w', encoding='utf8')
+        out = open(output_file, 'w', encoding='utf8')
    except OSError:
-        logger.error('Impossible to create file: %s', output_filename)
+        logger.error('Impossible to create file: %s', output_file)
        return None
    outfd = out.fileno()
@@ -529,6 +441,7 @@ def remux_srt_subtitles(mkvmerge_path:str, input_file: IO[bytes], output_filenam
    env = {**os.environ, 'LANG': 'C'}
    logger.info('Remux subtitles: %s', mkv_merge_params)
    with Popen(mkv_merge_params, stdout=PIPE, close_fds=False, env=env) as mkvmerge:
        assert mkvmerge.stdout is not None
        pb = tqdm(TextIOWrapper(mkvmerge.stdout, encoding="utf-8"), total=100, unit='%',
                  desc='Remux subtitles:')
        for line in pb:
@@ -537,8 +450,9 @@ def remux_srt_subtitles(mkvmerge_path:str, input_file: IO[bytes], output_filenam
                m = p.match(line)
                if m is None:
                    logger.error('Impossible to parse progress')
-                pb.n = int(m['progress'])
+                else:
-                pb.update()
+                    pb.n = int(m['progress'])
                    pb.update()
            elif line.startswith('Warning'):
                warnings.append(line)
@@ -551,82 +465,3 @@ def remux_srt_subtitles(mkvmerge_path:str, input_file: IO[bytes], output_filenam
            logger.error('Remux subtitles returns errors')
    return None
@typechecked
 def concatenate_h264_parts(h264parts: list[IO[bytes]], output: IO[bytes]) -> None:
    logger = logging.getLogger(__name__)
    total_length = 0
    for h264 in h264parts:
        fd = h264.fileno()
        total_length += fstat(fd).st_size
    logger.info('Total length: %d', total_length)
    outfd = output.fileno()
    lseek(outfd, 0, SEEK_SET)
    pb = tqdm(total=total_length, unit='bytes', desc='Concatenation')
    for h264 in h264parts:
        fd = h264.fileno()
        lseek(fd, 0, SEEK_SET)
        while True:
            buf = read(fd, 1000000)
            if buf is None or len(buf) == 0:
                break
            pos = 0
            while pos < len(buf):
                nb_bytes = write(outfd, buf[pos:])
                pb.update(nb_bytes)
                pos += nb_bytes
 def concatenate_h264_ts_parts(h264_ts_parts: list[IO[bytes]], output: IO[bytes]) -> None:
    logger = logging.getLogger(__name__)
    header = '# timestamp format v2\n'
    output.write(header)
    last = 0.
    first = True
    for part in h264_ts_parts:
        if first:
            offset = last
        else:
            # TODO: take framerate into account
            offset = last + 40
        logger.debug('Parsing file: %s. Offset=%d', part, offset)
        isheader = part.readline()
        if (not isheader) or (isheader != header):
            logger.error('Impossible to find a valid header: "%s"', isheader)
            exit(-1)
        while True:
            line = part.readline()
            if not line:
                break
            ts = offset + float(line)
            last = max(last,ts)
            output.write(f'{ts:f}\n')
        if first:
            first = False
 # TODO: finish this procedure
 def do_coarse_processing(ffmpeg_path:str, ffprobe_path:str, mkvmerge_path:str,
                         input_file: IO[bytes], begin, end, nb_frames, framerate,
                         files_prefix, streams, width, height, temporaries, dump_mem_fd) -> None:
    # pylint: disable=W0613
    logger = logging.getLogger(__name__)
    # Internal video with all streams (video, audio and subtitles)
    internal_mkv_name = f'{files_prefix}.mkv'
    try:
        internal_mkv = open(internal_mkv_name, 'wb+')
    except OSError:
        logger.error('Impossible to create file: %s', internal_mkv_name)
        exit(-1)
    # Extract internal part of MKV
    extract_mkv_part(mkvmerge_path=mkvmerge_path, input_file=input_file, output_file=internal_mkv,
                     begin=begin, end=end)
    temporaries.append(internal_mkv)
@@ -5,16 +5,19 @@
 import logging
 import re
 from io import BytesIO
 from math import ceil, log
 from os import write
 from typing import IO
 from io import BytesIO
 from typeguard import typechecked
 from tscut.exceptions import TemporaryFileError
 from tscut.temporaries import TemporaryFiles
@typechecked
-def dump_ppm(pictures: bytes, prefix: str, temporaries: list[IO[bytes]]) -> None:
+def dump_ppm(pictures: bytes, prefix: str, temporaries: TemporaryFiles) -> None:
    """
    Dump PPM pictures from a bytes buffer to files.
@@ -71,7 +74,7 @@ def dump_ppm(pictures: bytes, prefix: str, temporaries: list[IO[bytes]]) -> None
        header_len=2+1+ceil(log(width, 10))+1+ceil(log(height, 10))+1+3+1
        try:
            with open(filename, 'wb') as out:
-                temporaries.append(out)
+                temporaries.add(out)
                outfd = out.fileno()
                length=header_len+3*width*height
                nb_bytes = 0
@@ -79,5 +82,5 @@ def dump_ppm(pictures: bytes, prefix: str, temporaries: list[IO[bytes]]) -> None
                    nb_bytes+=write(outfd, pictures[pos+nb_bytes:pos+length])
                pos+=length
                picture+=1
-        except OSError:
+        except OSError as e:
-            logger.error('Impossible to create file: %s', filename)
+            raise TemporaryFileError(f"Impossible to create file {filename}") from e
@@ -9,6 +9,9 @@ from datetime import timedelta
 from typeguard import typechecked
 from tscut.exceptions import InvalidMediaError
 logger = logging.getLogger(__name__)
@typechecked
 def parse_timestamp(ts:str) -> timedelta|None:
@@ -29,8 +32,6 @@ def parse_timestamp(ts:str) -> timedelta|None:
                - the timestamp string is not in the correct format
                - the timestamp values are out of range (e.g. hour > 23, minute > 59, etc.)
    """
    logger = logging.getLogger(__name__)
    ts_reg_exp = (r'^(?P<hour>[0-9]{1,2}):(?P<minute>[0-9]{1,2})'
            r':(?P<second>[0-9]{1,2})(\.(?P<us>[0-9]{1,6}))?$')
    p = re.compile(ts_reg_exp)
@@ -65,9 +66,8 @@ def parse_timestamp(ts:str) -> timedelta|None:
    if us < 0 or us > 1000000:
        logger.error("milliseconds must be in [0,1000000[")
        return None
    res = timedelta(hours=hour, minutes=minute, seconds=second, microseconds=us)
-    return res
+    return timedelta(hours=hour, minutes=minute, seconds=second, microseconds=us)
@typechecked
 def parse_time_interval(interval: str) -> tuple[timedelta, timedelta] | tuple[None, None]:
@@ -90,8 +90,6 @@ def parse_time_interval(interval: str) -> tuple[timedelta, timedelta] | tuple[No
                - the time values are out of range (e.g. hour > 23, minute > 59, etc.)
                - the end time is before the start time (non-monotonic interval)
    """
    logger = logging.getLogger(__name__)
    interval_reg_exp = (r'^(?P<hour1>[0-9]{1,2}):(?P<minute1>[0-9]{1,2}):(?P<second1>[0-9]{1,2})'
                    r'(\.(?P<ms1>[0-9]{1,3}))?-(?P<hour2>[0-9]{1,2}):(?P<minute2>[0-9]{1,2})'
                    r':(?P<second2>[0-9]{1,2})(\.(?P<ms2>[0-9]{1,3}))?$')
@@ -193,30 +191,23 @@ def compare_time_interval(interval1: tuple[timedelta, timedelta],
@typechecked
-def get_ts_frame(frame: dict) -> timedelta|None:
+def get_ts_frame(frame: dict) -> timedelta:
    logger = logging.getLogger(__name__)
    if 'pts_time' in frame:
        pts_time = float(frame['pts_time'])
    elif 'pkt_pts_time' in frame:
        pts_time = float(frame['pkt_pts_time'])
    else:
-        logger.error('Impossible to find timestamp of frame %s', frame)
+        raise InvalidMediaError(f"Impossible to find timestamp of frame {frame}")
        return None
-    ts = timedelta(seconds=pts_time)
+    return timedelta(seconds=pts_time)
    return ts
@typechecked
-def get_packet_duration(packet: dict) -> int | None:
+def get_packet_duration(packet: dict) -> int:
    logger = logging.getLogger(__name__)
    if 'duration' in packet:
        duration = int(packet['duration'])
    elif 'pkt_duration' in packet:
        duration = int(packet['pkt_duration'])
    else:
-        logger.error('Impossible to find duration of packet %s', packet)
+        raise InvalidMediaError("Impossible to find duration of packet {packet}")
        return None
    return duration
Author	SHA1	Message	Date
Frédéric Tronel	814c64cc52	Fix a bug discovered by the synthetic video.	2026-08-31 21:28:07 +02:00
Frédéric Tronel	c17ccb6563	Typo in calling ffmpeg.	2026-08-31 21:27:41 +02:00
Frédéric Tronel	e1d7474e00	Improve error handling.	2026-08-31 14:30:38 +02:00
Frédéric Tronel	51bab76c00	Improve error handling.	2026-08-31 13:54:03 +02:00
Frédéric Tronel	153873637e	Improve temporaries management.	2026-08-31 12:00:12 +02:00
Frédéric Tronel	b0bc0a338e	Improve exception raising when it comes from an exception handler.	2026-08-31 11:20:25 +02:00
Frédéric Tronel	4597c459a8	Remove extraneous f-string.	2026-08-31 11:14:38 +02:00
Frédéric Tronel	5b2afd29b3	Add a dedicated class for temporary files management.	2026-08-31 11:12:47 +02:00
Frédéric Tronel	d1df44c82c	Remove exit from pipeline.	2026-08-31 08:57:28 +02:00
Frédéric Tronel	1a616f6cbd	Removing a lot of sys.exit inside cutting module.	2026-08-31 08:34:52 +02:00
Frédéric Tronel	d756fede54	H264 concatenation functions are moved to cutting module.	2026-08-30 17:45:03 +02:00
Frédéric Tronel	34d1bc26f3	Create a dedicated module for coarse cutting.	2026-08-30 17:34:57 +02:00
Frédéric Tronel	04781021eb	Create a dedicated module for coarse cutting.	2026-08-30 17:34:42 +02:00
Frédéric Tronel	e1ac3e48bb	Linting with ruff.	2026-08-30 17:24:26 +02:00
Frédéric Tronel	ae1a040590	Refactoring: extract cutting engine inside a dedicated module.	2026-08-30 17:19:03 +02:00
Frédéric Tronel	0345293664	More linting with ruff (suppress unused imports).	2026-08-30 14:45:25 +02:00
Frédéric Tronel	63d2deafb1	Linting with ruff.	2026-08-30 14:42:21 +02:00
Frédéric Tronel	6f5dd3a9e4	Final typing error.	2026-08-30 14:18:39 +02:00
Frédéric Tronel	093ad733d3	Even more typing errors fixes.	2026-08-30 14:14:18 +02:00
Frédéric Tronel	a6a41c112e	Fix of several typing errors.	2026-08-30 13:56:41 +02:00
Frédéric Tronel	447d04d13d	Better imports in the pipeline.	2026-08-30 09:33:02 +02:00
Frédéric Tronel	2590faf20f	The full pipeline is now in its own module.	2026-08-30 08:58:23 +02:00
Frédéric Tronel	bea482487e	Replace spurious exit call by raising exceptions.	2026-08-29 16:15:47 +02:00
Frédéric Tronel	c48b77f294	Move tests to root of the project.	2026-08-29 16:12:23 +02:00
Frédéric Tronel	0ff2e3a6dd	Linting and typing fixes.	2026-08-29 16:07:29 +02:00
Frédéric Tronel	294d857206	Put all exceptions in their dedicated module.	2026-08-29 15:30:47 +02:00
Frédéric Tronel	280eb1facb	Put all functions related to mkvtoolnix tools into a dedicated module.	2026-08-29 15:30:29 +02:00