`--depth` parameter silently ignored when using Y4M input files

[rachelplusplus]

Currently, when avifenc is run with an input file in .y4m format, the colour depth used in the resulting AVIF file is always the same as in the original Y4M file. This currently overrides any explicit bit depth specified on the command line.

While this behaviour is documented, it is non-intuitive. In addition, the --depth parameter is silently ignored, rather than printing an appropriate warning.

Ideally, if the input bit depth and --depth parameter differ, the input should be converted to the requested precision before compression. This can help squeeze a bit more out of the format: usually, even if the input file uses an 8-bit depth, compressing at a 10-bit internal precision gains a few % better compression at a negligible time cost, due to a reduction in rounding error within the codec.

If this is done, I would also like to request that avifenc normalize "unknown" colour space components to definite values. Otherwise, the colour space inferred by downstream tools can differ between 8-bit and 10-bit files, leading to the wrong colours being displayed.

[wantehchang]

Rachel: Thank you for the bug report. I will first make avifenc fail with an error if the .y4m input bit depth and the --depth parameter differ.

Could you clarify why the colour space inferred by downstream tools can differ between 8-bit and 10-bit AVIF files with "unknown" colour space components?

[juliobbv-p]

I will first make avifenc fail with an error if the .y4m input bit depth and the --depth parameter differ.

Another option to consider: we could treat y4ms like they were pngs, and treat --depth as if it were the "output" depth, so an 8-bit y4m can be up-converted to a 10-bit AVIF with --depth 10. y4ms are closer to pngs in usability than to raw yuvs, as all the width/height/bit depth info is included in them. Aligning y4m/png behavior feels like the least astonishing behavior.

Up-converting content from 8 to 10 bits can be noticeably beneficial for efficiency, as shown by the following comparison by Rachel herself: https://www.rachelplusplus.me.uk/blog/2025/07/a-better-image-compression-comparison/.