FFmpeg Multi Generation

[chris319]

Here is a straightforward if naive question about ffmpeg:

Is it possible to run a video clip through ffmpeg multiple times without image degradation, or can I expect the video to be degraded slightly with each pass?

I know that crf 0 is "lossless". Is there more to it than that? Will there be chroma subsampling or motion prediction?

For audio I would simply copy the existing audio track.

[LigH.de]

If you do not save intermediate results using a lossy compression format, or do not alter the video at all, you will not lose quality (up to possible chroma supsampling), but you also will get huge files.

CRF 0 only applies to MPEG video formats AVC/H.264 (or HEVC/H.265), but there are many more lossless video formats.

Which kind of editing do you intend, at all?

If it is multiple cutting, then be aware that the audio track may have different audio frame lengths compared to video frames, so you may introduce asynchronity at cut points if you keep the audio in a compressed format.

[chris319]

I'm thinking of processing raw camcorder video to begin with, so it's already H.264/mp4, and adding superimposed titles as the second step.

I could go camcorder mp4 -> level-corrected lossless -> lossless with titles. I'm thinking the two lossless steps won't degrade the h.264 any further.

What do you suggest as a lossless format?

[LigH.de]

In ffmpeg you have a wide choice with known efficient lossless compression, e.g.: FFV1, Ut Video, Huffyuv (ffvhuf), and also lossless AVC (I would prefer fixed quantizer 0 over CRF 0 here, but the result might be the same). In the MKV container you should be able to mix any of them with any audio track.

[pandy]

Perform all processing in single step or as already advised use lossless video codec as intermediate - i use FFV1 but is more oriented for archiving than real time signal flow. Huffyuv or its ffmpeg variant seem to be OK.