Yusuf Redžić

It seems like `estimate_inter_costs` actually needed more vertical padding (not horizontal padding which is what I thought originally), which is pretty easy to add without having to do a bunch...

I've fixed the standard scene detection mode, now I just need to address the other issues (mainly vapoursynth input support and downscaling for scene detection).

Surprisingly ffmpeg isn't actually the problem here, it's mostly vapoursynth which has a limited (at least, for our specific use case) decoding API. Vapoursynth seems to be based on the...

Update: I've added support in this PR for vapoursynth input. It seems to work on the scripts I've tried so far (even with HBD and `--sc-method standard`), but it might...

It looks like the standard scene detection mode with VS only happens to work with certain dimensions (but not with others, for example 930x734), so some kind of check will...

> Definitely possible, that's how we did it in AvsP. I have to go back and refresh my memory on exactly how that was done, but vital for colorspace conversion...

@se17p If av1an crashes with a segfault, it basically means that there is some kind of ABI mismatch for dynamically linked libraries. Make sure that vapoursynth and ffmpeg are both...

The cause of this problem appears to be related to https://github.com/rust-lang/rust/issues/45572, where the pipe buffer becomes full, blocking further output. The only problem I have with this approach is that...

@ls-milkyway Av1an itself does not really have minimum cpu requirements, the only thing is that our Windows binary assumes sse2 but you can compile Av1an with `-C target-feature=-sse2` if you...

Yeah, pretty much just echoing what @silverbacknet said here. `--ffmpeg` filters each chunk separately, and the filtering is not set up in a way such that it expects filters to...