Extracted clips don't have audio

[SKempin]

Can you confirm if the extracted clips should include audio? The clips are silent.

Thank you.

[Breakthrough]

Hi @SKempin;

That is correct. In the future I can consider adding a mode that uses an external tool (e.g. ffmpeg) to extract the audio as well.

Thanks!

[xiongyw]

It will be great to have such a feature. Thanks!

[Breakthrough]

The long-term approach to fixing this would be to have DVR-Scan rely on PySceneDetect for much of the heavy lifting (since most is already done in that project), however that's a more long-term goal to converge both projects (i.e. to rewrite DVR-Scan to use the PySceneDetect API).

In the meantime, you can always use the timecodes that DVR-Scan outputs in scan-only mode (-so) with ffmpeg to accomplish this manually or with a bash script. For those who are more Python-savvy, this can be accomplished using scenedetect.video_splitter.split_video_ffmpeg(...) (link).

For reference, currently videos are processed and encoded on the fly using OpenCV VideoWriter. Using ffmpeg instead for this I think is the right approach, and I think most usage of DVR-Scan is currently offline rather than online, indicating that this should be the preferred/default method. Another benefit would be improved video quality of the output (since two-pass encoding, or even just copying the codec if frame accuracy isn't a requirement), so this is definitely one of the highest priorities of all open issues for DVR-Scan.

I'll plan this for the next v1.3, although cannot comment on a definitive timeline just yet.

[Breakthrough]

In v1.5, there will be an option to use ffmpeg for generating output files. This allows you to both do re-encoding (-m ffmpeg) as well as codec copying (-m copy), e.g.:

dvr-scan -i video.mp4 -m ffmpeg
dvr-scan -i video.mp4 -m copy

Note that certain features will not work in this mode yet (mainly, any overlays, and concatenating multiple input videos). Support for these features with these modes can be considered if required by filing a new feature request.

Thanks!