Recherche avancée

Sur d’autres sites (12794)

• How to automate ffmpeg to split and merge parts of video, and keep the audio in sync ?

  9 décembre 2024, par Tree

  I have a Python script that automates trimming a large video (2 hours) into smaller segments and then concatenating them without re-encoding, to keep the process fast. The script runs these ffmpeg commands :

  


  import subprocess

# Extract chunks
segments = [(0, 300), (300, 600), (600, 900)]  # example segments in seconds
for i, (start, length) in enumerate(segments):
    subprocess.run([
        "ffmpeg", "-i", "input.mp4", "-ss", str(start), "-t", str(length),
        "-c", "copy", "-reset_timestamps", "1", "-y", f"chunk_{i}.mp4"
    ], check=True)

# Create concat list
with open("list.txt", "w") as f:
    for i in range(len(segments)):
        f.write(f"file 'chunk_{i}.mp4'\n")

# Concatenate
subprocess.run([
    "ffmpeg", "-f", "concat", "-safe", "0",
    "-i", "list.txt", "-c", "copy", "-y", "merged_output.mp4"
], check=True)


  


  All chunks come from the same source video, with identical codecs, resolution, and bitrate. Despite this, the final merged_output.mp4 sometimes has audio out of sync—especially after the first chunk.

  


  I’ve tried using -ss before -i to cut on keyframes, but the issue persists.

  


  Question : How can I ensure correct A/V sync in the final concatenated video when programmatically segmenting and merging via ffmpeg without fully re-encoding ? Is there a way to adjust the ffmpeg commands or process to avoid audio desynchronization ?

  


• Concatenate audio parts of different videos with ffmpeg [migrated]

  15 août 2024, par Swike

  So I have two videos :

  


  

  • Vid_eng.mkv is a high resolution video that has the audio in English.

  


  • Vid_esp.mkv is a low resolution video that has the audio in Spanish.

  


  


  Both videos are showing more or less the same stuff but the difference is that the Spanish one has the intro and outro cut out (the first 66 seconds are missing).

  


  I want to create a new video file, Vid_out.mkv that has the Vid_eng.mkv video and both audios (so the final file has multistream audio). The issue is that the Spanish audio will not be in sync with the high resolution video, so what I did was this :

  


  ffmpeg -i Vid_eng.mkv
       -f lavfi -t 66 -i anullsrc -i Vid_esp.mkv
       -filter_complex "[1:a][2:a]concat=n=2:v=0:a=1[outa]"
       -map 0 -map "[outa]"
       -metadata:s:a:0 language=eng
       -metadata:s:a:1 language=spa
       -disposition:a:1 default
       -c:v copy Vid_out.mkv


  


  What I'm doing here is essentially creating the new Vid_out.mkv using the video from Vid_eng.mkv, the audio multistream from both Vid_eng.mkv and Vid_esp.mkv, but with the twist that the audio from Vid_esp.mkv is actually beginning 66 seconds later (so that it is synced with the English video). I did this by adding 66 seconds of silence (with -f lavfi -t 66 -i anullsrc) and concatenating that silence with the Spanish audio (with -filter_complex "[1:a][2:a]concat=n=2:v=0:a=1[outa]") before adding this concatenated audio stream to the final file as the secondary audio. Also I included some metadata to know the which audio is which.

  


  Everything good for the moment. But now I wanted to try something more sophisticated : instead of having 66 seconds of silence concatenated to the Spanish audio I want 66 seconds of the English audio (the audio from Vid_eng.mkv), then the audio from Vid_esp.mkv, and finally when the Spanish audio has finished and the video is still going, I want the English audio again. So, instead of silence at the beginning and end of the video when played with the secondary audio stream, what I want is the English audio to play. How can I do that with ffmpeg ?

  


  ---

  


  For reference I've been following ideas from these answers :

  


  

  • How to add audios from different videos into one multistream video

  


  • How to add a few seconds of silence to an audio

  


  


• avcodec/vvcdec : split ctu table to zero init and no zero init parts

  28 juillet 2024, par Nuo Mi

  avcodec/vvcdec : split ctu table to zero init and no zero init parts
  cus need to init to zero, other parts are not
  

  • [DH] libavcodec/vvc/ctu.c
  • [DH] libavcodec/vvc/ctu.h
  • [DH] libavcodec/vvc/dec.c
  • [DH] libavcodec/vvc/dec.h
  • [DH] libavcodec/vvc/inter.c
  • [DH] libavcodec/vvc/intra.c