Dolby Vision now possible through MP4 Mux.

[RESET_9999]

The output wouldn't be easily playable if it sticks to lossless encoding, as the files would be huge.
And there are too many variables when reencoding lossy, so it's probably not worth it.

What did you use to convert the 16-bit RGB to YUV 10-bit without messing up the colors in the Rise of Gru P8(fel baked)sample?
Anything I tried in ffmpeg, I got a slight color shift compared to your sample: https://slow.pics/c/TiRTLS4F

I get quite different brightness reading as well: https://slow.pics/c/XQ7krMSf

[quietvoid]

What did you use to convert the 16-bit RGB to YUV 10-bit without messing up the colors in the Rise of Gru P8(fel baked)sample?
Anything I tried in ffmpeg, I got a slight color shift compared to your sample: https://slow.pics/c/TiRTLS4F

I get quite different brightness reading as well: https://slow.pics/c/XQ7krMSf

I didn't use DoViBaker, but you're supposed to specify 2020ncl for the target matrix when converting to YUV.

[danbez]

What did you use to convert the 16-bit RGB to YUV 10-bit without messing up the colors in the Rise of Gru P8(fel baked)sample?
Anything I tried in ffmpeg, I got a slight color shift compared to your sample: https://slow.pics/c/TiRTLS4F

I get quite different brightness reading as well: https://slow.pics/c/XQ7krMSf

I didn't use DoViBaker, but you're supposed to specify 2020ncl for the target matrix when converting to YUV.

I did a few DoViBaker conversions just to test the results. I used the command line below:
ffmpeg -i "example.avs" -x265-params "crf=14:hdr10-opt=1:chromaloc=2:range=limited:output-depth=10:profile=main10:level-idc=5.1:colorprim=bt2020:colormatrix=bt2020nc:transfer=smpte2084:hdr10=1:max-cll=0,0:master-display=G(13250,34500)B(7500,3000)R(34000,16000)WP(15635,16450)L(10000000,1)" "out10.hevc"

While I liked the results, I'm happy to get feedback on what can be improved.

[RESET_9999]

I use this cli...

%ffmpeg_path% -i "%filepath%%filename%.avs" -preset %preset% -x265-params crf=%crf%:repeat-headers=1:sar=1:hrd=1:hdr10-opt=1:aud=1:open-gop=0:hdr10=1:sao=0:rect=0:cutree=0:deblock=-3-3:strong-intra-smoothing=0:chromaloc=2:aq-mode=2:colorprim=bt2020:colormatrix=bt2020nc:transfer=smpte2084:vbv-maxrate=160000:vbv-bufsize=160000:max-luma=1023:max-cll=%maxcll_path%,%maxfall_path%:master-display=%MDP%WP(15635,16450)L(%max_pq%0000,%min_pq%) -pix_fmt yuv420p10le -y "%output_path%%filename%_HDR10.hevc"

What i dont understand is the hdr measurement difference. The avg_pq of "quietvoid sample is quite different than the original P7 rpu while the one created by this cmd is almost the same as the original.

[quietvoid]

What i dont understand is the hdr measurement difference. The avg_pq of "quietvoid sample is quite different than the original P7 rpu while the one created by this cmd is almost the same as the original.

Your measurements are off somehow. This is what I get from measuring my profile 8.1 encode: https://0x0.st/HQ26.png
And with the same scenes as original: https://0x0.st/HQ2l.png

[RESET_9999]

Your measurements are off somehow. This is what I get from measuring my profile 8.1 encode: https://0x0.st/HQ26.png
And with the same scenes as original: https://0x0.st/HQ2l.png

that is strange. I used madvr...
I figured out the color shift though. It needed ConvertToYUV420(matrix="2020") in the avs script. And now madvr read the avg_pq just like your original p8 sample

how did you measure it ?

[RESET_9999]

Ha, if I measure it in resolve instead, now it's closer to yours and almost identical to the original P7. So something is wrong with madvr measurement for this file, somehow. strange stuff...

https://slow.pics/c/GdrYzfzM

[quietvoid]

how did you measure it ?

I use my own script in VapourSynth.

[speeddemon]

Would it be possible to have a version of DoVi_Scripts someday that uses DoViBaker and dovi_tool to reconstruct the FEL image and output a correct P8.1 file?
@quietvoid & @RESET_9999 have either of you considered teaming up with erazortt to do this?
Thanks!

The output wouldn't be easily playable if it sticks to lossless encoding, as the files would be huge.
And there are too many variables when reencoding lossy, so it's probably not worth it.

@quietvoid and @RESET_9999 is this FEL expansion issue more of a problem for >1000-nit displays or is it also a problem on typical 700-1000-nit OLEDs?
Also, is there anywhere where one of you (or someone else) has completely explained why this FEL expansion issue exists and what is happening in detail?
Is it better to just watch in HDR instead of DV for movies with FEL expansion if a FEL compatible player isn't an option?

[RESET_9999]

@quietvoid and @RESET_9999 is this FEL expansion issue more of a problem for >1000-nit displays or is it also a problem on typical 700-1000-nit OLEDs?
Also, is there anywhere where one of you (or someone else) has completely explained why this FEL expansion issue exists and what is happening in detail?
Is it better to just watch in HDR instead of DV for movies with FEL expansion if a FEL compatible player isn't an option?

it's a problem for any display.
The TV receives brightness metadata (L1/L2) that is wrong for the content it's displaying so because it thinks the content is much brighter than it is, it dims the brightness accordingly to adapt to its capabilities.

see:
https://www.youtube.com/watch?v=nT1qBQLZIIc
https://www.youtube.com/watch?v=S0J0blslHjY
https://www.youtube.com/watch?v=VObky4peBXE

[danbez]

@quietvoid and @RESET_9999 is this FEL expansion issue more of a problem for >1000-nit displays or is it also a problem on typical 700-1000-nit OLEDs?
Also, is there anywhere where one of you (or someone else) has completely explained why this FEL expansion issue exists and what is happening in detail?
Is it better to just watch in HDR instead of DV for movies with FEL expansion if a FEL compatible player isn't an option?

it's a problem for any display.
The TV receives brightness metadata (L1/L2) that is wrong for the content it's displaying so because it thinks the content is much brighter than it is, it dims the brightness accordingly to adapt to its capabilities.

see:
https://www.youtube.com/watch?v=nT1qBQLZIIc
https://www.youtube.com/watch?v=S0J0blslHjY
https://www.youtube.com/watch?v=VObky4peBXE

On top of that, we also have titles where the Base Layer was poorly authored resulting in a lot of compression artifacts and other issues that are only fixed with the FEL layer. Typical examples are Total Recall, Halloween 2 and a few others.

[speeddemon]

The TV receives brightness metadata (L1/L2) that is wrong for the content it's displaying so because it thinks the content is much brighter than it is, it dims the brightness accordingly to adapt to its capabilities.

Do you/we have any idea why they're doing this?
Is it better to just watch in HDR instead of DV for movies with FEL expansion if a FEL compatible player isn't an option?
Would it be possible to re-write the L1/L2 metadata so it's corrected/"unexpanded"?

[RESET_9999]

Do you/we have any idea why they're doing this?
Is it better to just watch in HDR instead of DV for movies with FEL expansion if a FEL compatible player isn't an option?
Would it be possible to re-write the L1/L2 metadata so it's corrected/"unexpanded"?

4 options:
1- just watch in HDR10
2- bake the EL into the BL
3- generate a proper RPU in resolve for the BL actual brightness.
4- get a proper player

Expanding the brightness makes sense actually. It futureproofs the grade for when the TV will be brighter and it makes the content accurate on more type(brightness, HDR format) of display.

Because this way, people with brighter DV TV(G3) can enjoy brighter HDR and people with low nits DV TV can also get an accurate picture thanks to the trim pass and L1.
Then, people with HDR10-only TVs (low nits) can also get an accurate image with the 1000nits BL. If the BL was as bright as the EL, then those low-nits HDR10 TVs would look bad/clipped.

On top of that, we also have titles where the Base Layer was poorly authored resulting in a lot of compression artifacts and other issues that are only fixed with the FEL layer. Typical examples are Total Recall, Halloween 2 and a few others.

That's right. I found a couple that improve grain/details: https://docs.google.com/spreadsheets/d/ ... =917134052

[speeddemon]

3- generate a proper RPU in resolve for the BL actual brightness.

Do you have a tutorial for this or know of a tutorial for this?

Thanks!

[RESET_9999]

Do you have a tutorial for this or know of a tutorial for this?
Thanks!

https://www.youtube.com/watch?v=ZdFiqATo_j8