ynput
Deep images fail with ffmpeg in `ExtractReview`
Is there an existing issue for this?
- [x] I have searched the existing issues
Current Behavior:
Exporting Deep data is now allowed by https://github.com/ynput/ayon-houdini/pull/120
Exporting Deep Beauty AOV fails with ffmpeg in ExtractReview, it seems that the current ffmpeg didn't like the deep image.
Not sure if this issue related to core addon or third party addon.
[exr @ 000001d9fab74940] deep data is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
[image2 @ 000001d9fab71f00] decoding for stream 0 failed
[image2 @ 000001d9fab71f00] Could not find codec parameters for stream 0 (Video: exr, none): unspecified size
Consider increasing the value for the 'analyzeduration' (0) and 'probesize' (5000000) options
Input #0, image2, from 'H:/AYON/projects/Experiments/things/USD_Render/work/Houdini/renders/Expr_USD_Render_Houdini_v006/karmarendersettings_and_beauty_products\Expr_USD_Render_Houdini_v006.karmarendersettings_and_beauty_products.dcm.%d.exr':
Duration: 00:00:00.08, start: 0.000000, bitrate: N/A
Stream #0:0: Video: exr, none, 25 fps, 25 tbr, 25 tbn, 25 tbc
Codec AVOption b (set bitrate (in bits/s)) specified for output file #0 (H:/AYON/projects/Experiments/things/USD_Render/work/Houdini/renders/Expr_USD_Render_Houdini_v006/karmarendersettings_and_beauty_products\Expr_USD_Render_Houdini_v006.karmarendersettings_and_beauty_products.dcm_h264.mp4) has not been used for any stream. The most likely reason is either wrong type (e.g. a video option with no video streams) or that it is a private option of some encoder which was not actually used for any stream.
Stream mapping:
Stream #0:0 -> #0:0 (exr (native) -> h264 (libx264))
Press [q] to stop, [?] for help
[exr @ 000001d9fabbd0c0] deep data is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
Error while decoding stream #0:0: Not yet implemented in FFmpeg, patches welcome
[exr @ 000001d9fb612d80] deep data is not implemented. Update your FFmpeg version to the newest one from Git. If the problem still occurs, it means that your file has a feature which has not been implemented.
Error while decoding stream #0:0: Not yet implemented in FFmpeg, patches welcome
Cannot determine format of input stream 0:0 after EOF
Error marking filters as finished
Conversion failed!
Expected Behavior:
It should work.
Version
1.0.0
What platform you are running on?
Windows
Steps To Reproduce:
- Setup Render in Houdini with deep beauty pass enabled in USD karma render.
- Publish
Are there any labels you wish to add?
- [x] I have added the relevant labels to the bug report.
Relevant log output:
Additional context:
No response
@MustafaJafar Can you try with the latest ffmpeg 7.1 ? Not sure this has been implemented: deep images use a different OpenEXR API and need to be flattened before being processed by ffmpeg.
I looked at the ffmpeg changelog and deep support is not mentioned.
The simplest solution is to pre-process the files with oiiotool deepRender.exr -flatten -o flatRender.exr.
Definitively not implemented in latest ffmpeg ! 😂
if (flags & 0x08) {
avpriv_report_missing_feature(s->avctx, "deep data");
return AVERROR_PATCHWELCOME;
}
I wonder whether we could workaround it for new by running it via ExtractOIIOTranscode plugin and convert it to something more FFMPEG friendly.
UPDATE ------ It actually is not working either see the internal comunication. https://discord.com/channels/517362899170230292/1344323811218034759
Executing arguments was not successful: "(['C:\\Users\\deadline.user\\AppData\\Local\\Ynput\\AYON\\addons\\ayon_third_party_1.1.1\\ayon_third_party\\downloads\\oiio\\oiiotool.exe', '--nosoftwareattrib', '--colorconfig', '//192.168.80.52/central_software/OCIO_2.1/studio-config-v1.0.0_aces-v1.3_ocio-v2.1.ocio', '-i:ch=R,G,B,A', '\\\\############\\projects/#######/scenes/###/###_0160/work/mist_P048/render/AWG_0160_mist_P048_v006/KMA_cbs_cubes\\###_0160_mist_P048_v006.KMA_cbs_cubes.dcm.989-992#.exr', '--ch', 'R=R,G=G,B=B,A=A', '--iscolorspace', 'ACEScg', '--ociodisplay:subimages=0', 'sRGB - Display', 'ACES 1.0 - SDR Video', '-o', 'C:\\Users\\DEADLI~1.USE\\AppData\\Local\\Temp\\ay_tmp_8l6iqbab\\####_0160_mist_P048_v006.KMA_cbs_cubes.dcm.989-992#.png'],)
I wonder whether we could workaround it for new by running it via
ExtractOIIOTranscodeplugin and convert it to something more FFMPEG friendly.UPDATE ------ It actually is not working either see the internal comunication. https://discord.com/channels/517362899170230292/1344323811218034759
Executing arguments was not successful: "(['C:\\Users\\deadline.user\\AppData\\Local\\Ynput\\AYON\\addons\\ayon_third_party_1.1.1\\ayon_third_party\\downloads\\oiio\\oiiotool.exe', '--nosoftwareattrib', '--colorconfig', '//192.168.80.52/central_software/OCIO_2.1/studio-config-v1.0.0_aces-v1.3_ocio-v2.1.ocio', '-i:ch=R,G,B,A', '\\\\############\\projects/#######/scenes/###/###_0160/work/mist_P048/render/AWG_0160_mist_P048_v006/KMA_cbs_cubes\\###_0160_mist_P048_v006.KMA_cbs_cubes.dcm.989-992#.exr', '--ch', 'R=R,G=G,B=B,A=A', '--iscolorspace', 'ACEScg', '--ociodisplay:subimages=0', 'sRGB - Display', 'ACES 1.0 - SDR Video', '-o', 'C:\\Users\\DEADLI~1.USE\\AppData\\Local\\Temp\\ay_tmp_8l6iqbab\\####_0160_mist_P048_v006.KMA_cbs_cubes.dcm.989-992#.png'],)
I wonder if newer versions of oiio could solve this - because it does have a full section on "Deep EXRs" in their docs: https://openimageio.readthedocs.io/en/latest/imageoutput.html#writing-deep-data
Does anyone have an example dcm file to share? If so, then we can just try it quickly with uv:
uv run --with OpenImageIO -- oiiotool.exe
Crash may be fixed with newer OIIO
Third party addon currently comes with OpenImageIO 2.3.10
The latest OpenImageIO (that's available also through pip now) is OpenImageIO 3.0.3.1
Relevant changes since then may be:
- fix(exr): Avoid integer overflow for large deep exr slice strides PR (3.0.1.0) in 3.0.1.0
- exr: Avoid integer overflow for large deep exr slice strides PR in 2.5.18.0
- IBAprep should not zero out deep images when creating dst PR in 2.4.8.0
- Fix writing deep exrs when buffer datatype doesn't match the file. PR in 2.3.14.0
I gave this a go and this fails:
uv run --with OpenImageIO -- oiiotool.exe -i:ch=R,G,B,A %FILE% -o test.png
Due to error:
oiiotool ERROR: -o : png does not support 'deep' images
But this succeeds:
uv run --with OpenImageIO -- oiiotool.exe -i:ch=R,G,B,A %FILE% --flatten -o test.png
As such, it may just be sufficient to include the --flatten argument for extract transcode settings to png and jpeg?
For example:
This test was running OpenImageIO 3.0.3.1
Also, here's an example DCM file:
With courtesy from @MustafaJafar