convert h.264 to lagarith/utvideo/huffyuv

[nixiejames]

are there software that don't downsample h.264 files and straight convert them to lossless codecs like ut video?

[sneaker]

ffmpeg can encode from H.264 to a number of lossless formats (including lossless H.264). You do know file size will usually multiply when converting a lossy format to a lossless one, don't you? There's no way around it in case you had that in mind when you said "don't downsample".

[nixiejames]

yes, allow me to explain. i want to import this h.264 file into premiere but in my experience premiere always downsamples h.264 files before you even get to edit/export. i thought i'd convert it to a lossless format first before premiere.

[Gramps]

Ffmpeg can encode to utvideo and huffyuv, but not lagarith. The ffmpeg developers were able to replicate the decoder portion of lagarith, but not the encoder.

[nixiejames]

can you guys recommend a good gui that can output to utvideo/huffyuv? i'm not used to cl and i don't prioritize learning it as of now.

[Gramps]

Both Xmedia Recode and AVIDemux have built in support for encoding Huffyuv in an AVI container. UTvideo can only be done via the FFMpeg commandline or, if you have installed the VFW UTVideo codec, via Virtualdub. Someone else might know of a GUI encoder that supports it OOTB, If you do go with UTVideo, you will need to install the codec anyway so that Premiere can read the file.

Word to the wise, if you install it, get the individual codec from the Software Codecs section on this site. Don't install it via one of those awful codec packs.

Note that files encoded using UTVideo tend to be quite a bit smaller than Huffyuv, if space is an issue at all.

[Dougster]

Selur's Hybrid can convert to both UTvideo and Huffyuv.

[Gramps]

I know that you can also use WinFF to encode UTvideo and HuffYUV too, but you have to create custom presets to do it.

[nixiejames]

i tried xmedia recode. my source is h.264 in yuv 4:4:4 which is not available in the software's version of huffyuv so i tried ffv1. looks like there's always quality loss even if it's a lossless format as long as it's a different format.

[jagabo]

looks like there's always quality loss even if it's a lossless format as long as it's a different format.

In converting 4:4:4 to 4:2:0 the color information is resized to half (both dimensions) the size of the luma. For example a 1920x1080 4:4:4 video has a 1920x1080 luma plane and two 1920x1080 chroma planes. After conversion to 4:2:0 it has a 1920x1080 luma plane and two 960x540 chroma planes. So of course there's a loss of quality. The chroma subsampling causes losses before the compression codec received it.

Keep in mind that all distribution formats (DVD, Blu-ray, broadcast TV, Youtube, etc.) use 4:2:0 subampling.

[nixiejames]

i meant to say every different codec whether it's in the same color space or not. the one i tried in xmedia recode was ffv1 in 4:4:4. there was still quality loss so that's where i got the idea.

[poisondeathray]

i meant to say every different codec whether it's in the same color space or not. the one i tried in xmedia recode was ffv1 in 4:4:4. there was still quality loss so that's where i got the idea.

There is no quality loss with ffv1.

It might be the method you used, or your decoding method, or host application isn't handling it properly

[jagabo]

i meant to say every different codec whether it's in the same color space or not. the one i tried in xmedia recode was ffv1 in 4:4:4. there was still quality loss so that's where i got the idea.

Lossless codecs are called lossless for a reason. You did something wrong. Or the program did some conversions internally.

[nixiejames]

first i captured gameplay using obs. these were my settings:
color format: i444
yuv color space: yuv
yuv color range: full

container: avi
video encoder: h264_nvenc
video encoder settings: preset=losslesshp

and there i got my video. i converted it to ffv1 via xmedia recode using default settings as i thought there were no settings needed tinkering.

i just compared them using snapshots from vlc though surely there are better ways you can suggest.

[poisondeathray]

first i captured gameplay using obs. these were my settings:
color format: i444
yuv color space: yuv
yuv color range: full

container: avi
video encoder: h264_nvenc
video encoder settings: preset=losslesshp

and there i got my video. i converted it to ffv1 via xmedia recode using default settings as i thought there were no settings needed tinkering.

i just compared them using snapshots from vlc though surely there are better ways you can suggest.

If you used full range YUV444, that is already lossy because the gameplay source is RGB. If you want truly lossless, record in RGB. What is the reason for using FFV1 or other lossless intermediate ? Another reason not to use YUV444 and full range is most editing programs cannot handle YUV444 properly, but most can handle RGB.

OBS will do a conversion to YUV444. That conversion can be done in a number of ways. IF you specified full range, it might use full range 601 or full range 709 . In order to "see" it properly, you have to use the same (reverse transform) method to convert it back to RGB for display. So likely VLC isn't using the proper viewing method (probably using limited range 709)

You might have to specify -pix_fmt yuv444p for 8bit YUV444, otherwise the xmediarecode might be doing other conversions.
-pix_fmt yuvj444p specifies full range 8bit YUV444 (the "j") , but it's not necessary, you just need to make sure the receiving program handles it properly

Also, apparently there were problems with OBS and NVENC lossless in the past, I don't know if they fixed things yet. But doing everything with ffmpeg works with nvenc lossless, and viewing it with full range 709 (reverse transform)

[nixiejames]

I forgot to say it's using full range 709. And yes, I know the source is RGB, but I think that's the most I can get from nvenc. It can't do rgb. I usually record with utvideo rgb using dxtory but now I'm trying nvenc because it's easier on cpu.

When I converted using xmedia recode, I set the color space to automatic. The resulting file was in YUV444. I believe using screenshots was a bad idea. Most players I've encountered have a hard time decoding compressed h264 files encoded using nvidia.

[raffriff42]

Recording in full range may be the problem, as most software expects YUV to be limited range unless instructed otherwise. Your final output will be limited range, so you might as well capture in limited range IMHO - but if you choose full range, don't use automatic color settings.