Turn on Audio Sync for audio extraction process

[dvh312]

Problem description

• I used Whisper AI and recognized that the subtitles generated didn't have the correct timestamps, especially near the end of the video.

Cause

1. The video files I'm using have some corrupted frames.
2. ffmpeg try to "skip" these frames, causing the duration of the extracted audio to be significantly shorter than the original video duration.

Solution

• Turning on audio sync to make the extracted audio matches the video timestamps.

Screenshots

Original command

ffmpeg  -i input.mp4  -acodec pcm_s16le -ac 1 -ar 16000 -f s16le output.wav 

ffmpeg-before.log

Imported to Audacity, only shows the duration of 22m52s532ms. The video duration is 22m59s960ms. This means subtitles will be shifted by almost 7 seconds ahead.

Updated command

ffmpeg  -i input.mp4  -acodec pcm_s16le -ac 1 -ar 16000 -f s16le -af aresample=async=1 output.wav 

ffmpeg-after.log

Audacity now shows the duration of 22m59s349ms, which is approximately the original video duration of 22m59s960ms

[morpheus65535]

You can't patch files in libs directory since they'll get overwritten when update update dependencies. You must find a way to do it from the provider itself or make a PR to ffprobe module itself, get it merged and we'll get it in next dependencies update cycle.

[anderson-oki]

@dvh312 Morpheus means you can't, there is a typo.

[dvh312]

@morpheus65535 Thanks, let me find a way to do it in the provider.

[dvh312]

@morpheus65535 I've reverted the change in libs and will make the update to ffsubsync directly.