Colorspace Conversions - YUY2 and YV12

[Christina]

With regards to SD home videos captured lossless YUY2 with HuffYUV codec with target delivery codec h264/mp4:

If all of my AviSynth filters work in YUY2, should I still be converting to YV12 anyway within my script?

If so, should I still do it at the top of my script? Or as the very last thing I do?

i.e. Is there any benefit in retaining YUY2 as long as possible or better to convert to YV12 early on before all the filters? Is there any speed gain if converting to YV12 before QTGMC? My version of QTGMC accepts YUY2.

Or, should I leave it YUY2 and just add -pix_fmt yuv420p to my ffmpeg batch file script when converting to h264?


2nd Question:

If my target output format is instead ProRes 422, do I just keep it in YUY2 the whole time? I assume if I convert to YV12 in the Avisynth script, I have to convert back in the ffmpeg conversion to ProRes because ProRes is 422, not 420?

[jagabo]

With regards to SD home videos captured lossless YUY2 with HuffYUV codec with target delivery codec h264/mp4:

If all of my AviSynth filters work in YUY2, should I still be converting to YV12 anyway within my script?

Do you plan to play your video on anything other than a computer? Pretty much nothing else plays 4:2:2 AVC. I would do it in my AviSynth script, near the end. Let the earlier filters work with as much data as possible.

If my target output format is instead ProRes 422, do I just keep it in YUY2 the whole time?

Yes.

[Christina]

Pretty much nothing else plays 4:2:2 AVC. I would do it in my AviSynth script, near the end. Let the earlier filters work with as much data as possible.

Thanks for responding so quickly.

Is there any difference in doing it at the end of the AviSynth script vs adding -pix_fmt yuv420p to my ffmpeg script when converting to h264? Or are they doing the same exact thing?

And just simply out of curiosity, does QTGMC take any longer in YUY2 because there is more color data stored in YUY2 compared with YV12?

[jagabo]

Is there any difference in doing it at the end of the AviSynth script vs adding -pix_fmt yuv420p to my ffmpeg script when converting to h264? Or are they doing the same exact thing?

They are doing the same basic thing but I don't know if they do it exactly the same way. Note that there are different chroma placements and scaling algorithms. I know AviSynth better than ffmpeg so I prefer to do everything in AviSynth. ffmpeg may have all the same options but I don't know the syntax.

And just simply out of curiosity, does QTGMC take any longer in YUY2 because there is more color data stored in YUY2 compared with YV12?

I don't know, I've never compared them. But other issues will come into play: the time it takes to do the conversion, the amount of work handling the different data layouts, the speed of the codec with the two formats (assuming you're compressing the output).

[poisondeathray]

Both ffmpeg and avisynth use "bicubic" for the default scaling algorithm (that includes the chroma scaling), but their implementation of bicubic are slightly different. It's a negligible difference unless you pixel peep. You can use different scaling algorithms in ffmpeg by controlling the sws flags, or by using zscale instead. In avisynth you could use ConvertToYV12(chromasample="blah") for the controlling the chroma downscaling algorithm

Yes, YV12 is faster than YUY2 for processing. Also, "planar" formats are generally faster for processing in avisynth than "packed" formats . e.g. YV16 is 422 "planar", YUY2 is 422 "packed" or interleaved

If you're not using prores, then converting to YV12 earlier on is beneficial for speed