Lossless Extracted Frames

[Jay123210599]

How do I extract frames from videos into lossless avif, heic/heif, or any other lossless formats (not tiff) that support YUV (4:2:0, 4:2:2, and 4:4:4)? From both H.264 and H.265 videos?

Are there any guis for these? How do I use them and/or ffmpeg to make image sequences of those files?
https://github.com/AOMediaCodec/libavif
https://github.com/strukturag/libheif

Also, when making jxl files, I noticed that making them regularly and making them with the special commands to make them lossless look the same. Is that true? How do I set the primaries and transfer needed to fix the error I mentioned in a previous post and hopefully make the process do faster?

[_Al_]

I don't know, but want to say after monitoring your questions in recent days, that you actually might start to look into Avisynth or Vapoursynth, meaning read their manuals, what they do, like conversions from format to other format the best way possible, there is also imwri plugin available to write rgb 8bit to disk. As pdr mentioned, it does not matter what video format you have, as humans, we can only observe it on screen as RGB 8 bit (regular monitors). Are you aware what Avisynth or Vapoursynth can do?

There are filters available that are temporal, not treating video as images (spatial only), filters work in yuv or rgb. And if your processes need to be automatized, I'd recommend vapoursynth (using python). You still did not reveal why you need to get yuv images from your video.

[poisondeathray]

Also, when making jxl files, I noticed that making them regularly and making them with the special commands to make them lossless look the same. Is that true?

No, it was explained in your other post and the help file. READ IT!

-distance 0 is lossless. -distance > 0 is not . Lower values are higher in quality for lossy mode

Encoder libjxl [libjxl JPEG XL]:
General capabilities: threads
Threading capabilities: other
Supported pixel formats: rgb24 rgba rgb48le rgba64le gray ya8 gray16le ya16le grayf32le
libjxl AVOptions:
-effort <int> E..V....... Encoding effort (from 1 to 9) (default 7)
-distance <float> E..V....... Maximum Butteraugli distance (quality setting, lower = better, zero = lossless, default 1.0) (from -1 to 15) (default -1)
-modular <int> E..V....... Force modular mode (from 0 to 1) (default 0)

How do I set the primaries and transfer needed to fix the error I mentioned in a previous post

-color_primaries bt709 -color_trc iec61966-2-1

and hopefully make the process do faster?

Again, it's "effort". Higher values are slower, produce smaller filesizes . If you want faster, filesizes for lossless will be larger, use a lower effort number

-effort <int> E..V....... Encoding effort (from 1 to 9) (default 7)

[pandy]

You can extract video frames exactly same way as in case of previously asked PNG/APNG. If you wish to use different video codec then use it instead PNG/APNG - ffmpeg command line stays similar - difference will be related to selected video codec.

[Jay123210599]

Also, when making jxl files, I noticed that making them regularly and making them with the special commands to make them lossless look the same. Is that true?

No, it was explained in your other post and the help file. READ IT!

-distance 0 is lossless. -distance > 0 is not . Lower values are higher in quality for lossy mode

Encoder libjxl [libjxl JPEG XL]:
General capabilities: threads
Threading capabilities: other
Supported pixel formats: rgb24 rgba rgb48le rgba64le gray ya8 gray16le ya16le grayf32le
libjxl AVOptions:
-effort <int> E..V....... Encoding effort (from 1 to 9) (default 7)
-distance <float> E..V....... Maximum Butteraugli distance (quality setting, lower = better, zero = lossless, default 1.0) (from -1 to 15) (default -1)
-modular <int> E..V....... Force modular mode (from 0 to 1) (default 0)

How do I set the primaries and transfer needed to fix the error I mentioned in a previous post

-color_primaries bt709 -color_trc iec61966-2-1

and hopefully make the process do faster?

Again, it's "effort". Higher values are slower, produce smaller filesizes . If you want faster, filesizes for lossless will be larger, use a lower effort number

-effort <int> E..V....... Encoding effort (from 1 to 9) (default 7)

What's -color_trc iec61966-2-1?

[poisondeathray]

What's -color_trc iec61966-2-1?

sRGB transfer as described in the IEC 61966-2-1:1999 document
https://en.wikipedia.org/wiki/SRGB

RGB is a standard RGB (red, green, blue) color space that HP and Microsoft created cooperatively in 1996 to use on monitors, printers, and the World Wide Web.[2] It was subsequently standardized by the International Electrotechnical Commission (IEC) as IEC 61966-2-1:1999

Check Table E.4 in the h.265 document for commonly used transfer characteristics , including sRGB

[Jay123210599]

What's -color_trc iec61966-2-1?

sRGB transfer as described in the IEC 61966-2-1:1999 document
https://en.wikipedia.org/wiki/SRGB

RGB is a standard RGB (red, green, blue) color space that HP and Microsoft created cooperatively in 1996 to use on monitors, printers, and the World Wide Web.[2] It was subsequently standardized by the International Electrotechnical Commission (IEC) as IEC 61966-2-1:1999

Check Table E.4 in the h.265 document for commonly used transfer characteristics , including sRGB

Can you tell the difference between these two jxl files?

[poisondeathray]

Can you tell the difference between these two jxl files?

Yes

[Jay123210599]

Can you tell the difference between these two jxl files?

Yes

Show me.

[poisondeathray]

Can you tell the difference between these two jxl files?

Yes

Show me.

Zoom in with an image editor

[Jay123210599]

Can you tell the difference between these two jxl files?

Yes

Show me.

Zoom in with an image editor

But it was visually lossless, right?

[poisondeathray]

But it was visually lossless, right?

I wouldn't call it visually lossless, but it might be for some people. There are larger differences between these two , than the differences between other two in the other example

[Jay123210599]

But it was visually lossless, right?

I wouldn't call it visually lossless, but it might be for some people. There are larger differences between these two , than the differences between other two in the other example

Are there lossless formats that have the same file size as the lossy jxl, or at least close to it?

[poisondeathray]

Are there lossless formats that have the same file size as the lossy jxl, or at least close to it?

No

Anything lossless will have substantially larger file sizes

Also any image format will be larger than the equivalent compression for video, because there is no benefit temporal compression for that image format .

[_Al_]

Can you tell the difference between these two jxl files?

Comparison of details between those two images,
click on each image about three time to get full size into a browser tab, then clicking those tabs you see differences. Each little square represents actual pixel.

[Jay123210599]

@poisondeathray and @_Al_

How about now? And let's be honest, nobody's going to bother zoom in on images to check for small differences these days, right?

[poisondeathray]

How about now? And let's be honest, nobody's going to bother zoom in on images to check for small differences these days, right?

Yes you can see differences.

But vast majority of general public won't care or notice.

It depends on why you are posting it, where you post it, and who your audience is.

Honestly - Die hard anime fans tend to notice those types of things. Especially edge and line noise, textures (noise or loss). They are probably among the pickiest videophiles