Versatile Video Coding (VVC) / H.266: HEVC successor

[Z2697]

Originally Posted by FranceBB

More interesting stuff coming from Intel with the two following commits to lavc

1)
https://github.com/FFmpeg/FFmpeg/com...c5eca35f5dfbfd

2) https://github.com/FFmpeg/FFmpeg/com...e6673ab8f66bf4

They have now exposed hardware decoding APIs in lavc so that even on Linux their new Lunar Lake GPUs can have hardware accelerated decoding for H.266 VVC. This will again spread soon via ffmpeg to things like MPV and it will make H.266 decoding easier. I know that Intel isn't exactly in a good place right now, but it's nice to see them directly contributing to open source projects. Also we're still at the end of 2024, it's a long way before July 2026, but it's nice to see widespread software decoding support thanks to libav and now the first cross platform hardware decoding support.

What's July 2026 about?

[FranceBB]

Originally Posted by Z2697

What's July 2026 about?

TX, i.e beginning of linear channel tv transmissions in H.266 VVC to end consumer devices.

[Z2697]

Originally Posted by FranceBB

TX, i.e beginning of linear channel tv transmissions in H.266 VVC to end consumer devices.

Long way to go

[kurkosdr]

Originally Posted by FranceBB

TX, i.e beginning of linear channel tv transmissions in H.266 VVC to end consumer devices.

Source? And what countries?

[FranceBB]

Originally Posted by kurkosdr

Source? And what countries?

Nothing new here, I'm still talking about Brazil.
I don't live there, but I'm keeping an eye on TV 3.0 and what the V-Nova guys post on LinkedIn 'cause they're gonna be at the forefront of this, so if they succeed then H.266 more generally will succeed in broadcast.

The timeline stayed the same so far by the way, but I've added the new hardware decoding support that Intel just submitted to lavc:

- July 2020 -> H.266 VVC specs finalized
- October 2020 -> VVEnc & VVDec by Fraunhofer are released
- July 2021 -> First chipset with hardware decoding released by MediaTek
- July 2022 -> First TV prototype shown at CES
- April 2024 -> Libav adds software decoding support (FFMpeg, VLC, MPV, Avisynth, VapourSynth)
- July 2024 -> Consumer TV becoming available
- September 2024 -> Intel Lunar Lake Graphics hardware decoding support
- November 2024 -> Lavc adds hardware decoding support for Intel Lunar Lake Graphics (FFmpeg, VLC, MPV)


Which leads us to:

- 2025 -> Initial H.266 VVC broadcasting tests in Brazil
- 2026 -> Beginning of H.266 VVC transmissions in Brazil

[Selur]

Does any of the vvc encoders atm. support other color sampling than 4:2:0 ?

[LigH]

New uploads: [Windows][GCC 14.2.0][64 bit]

Fraunhofer VVC Encoder ver. 1.12.1-rc1 da540e6

Fraunhofer VVC Decoder ver. 3.0.0 a646129

[birdie]

Originally Posted by Selur

Does any of the vvc encoders atm. support other color sampling than 4:2:0 ?

The second link in Google and https://en.wikipedia.org/wiki/Versatile_Video_Coding

Have you even tried?

https://vcgit.hhi.fraunhofer.de/jvet/VVCSoftware_VTM

YUV420
YUV422
YUV444

and 12bit encoding (probably)

It's dead slow.

[Z2697]

If you take "4:0:0" into account... hmm... both VVenC and uvg266 seem to support that.
And while VVenC have options related to 4:4:4 and 4:2:2 in the command line help message, it throws error when I give it YUV444 Y4M, I think it's not implemented yet.

[LigH]

VTM is the "reference software", hardly optimized at all. If any encoder is "complete", then this one. But not practically useable.

PS: You can try vvencFFapp:

        --InputChromaFormat [420]    input file chroma format (400, 420, 422, 444)

Please note that the options are so verbose, you better create a configuration file instead of trying to feed all options via CLI.

[Selur]

VTM is the "reference software", hardly optimized at all. If any encoder is "complete", then this one. But not practically useable.

I agree, VTM is not useable.

vvencapp states:

  -c,   --format [yuv420]            set input format (yuv420, yuv420_10, yuv420_10_packed, yuv400 (gray), yuv400_10

uvg266 states:

--input-format <string> : P420 or P400 [P420]

=> I will look at vvencFFapp then. (ah, only the full help does state InputChromaFormat )

Thanks!

Cu Selur

[Sagittaire]

Originally Posted by LigH

VTM is the "reference software", hardly optimized at all. If any encoder is "complete", then this one. But not practically useable.

Well if you use x265 with preset placebo or SVT-AV1 with preset 0, you obtain really low speed too. Use the slowest preset for codec is never obligation.

VVCEnc with preset "faster" will produce the same speed than "veryslow" for x264, "slow" for x265 or "complexity 5" for SVT-AV1.

Moreover VVCEnc is not really optimized for more than 12 thread. If you want make encoding at 100% CPU charge for 32 thread CPU and more, you must use multipart encoding.

For exemple for VVCEnc in fast mode preset you obtain 20 fps for 1080p10 for 7950X CPU and 40 fps for preset faster.

[Jamaika]

Originally Posted by Selur

Does any of the vvc encoders atm. support other color sampling than 4:2:0 ?

Try this: https://www.sendspace.com/file/tia55v

ffmpeg_avx2.exe -v verbose -f concat -safe 0 -i mylist_v.txt -f concat -safe 0 -i mylist_a.txt -c:v libvvenc -vb 6000k -c:a aac -ac 2 -ar 48000 -ab 128k -frames:v 100 -movflags faststart -threads 0 -vf scale=1920:1080:in_range=full: out_range=full,format=yuv422p10le 112.vvc
vvencapp_avx2.exe -i 111.y4m -s 1920x1080 -c yuv422 -b 6M -ip 256 --passes 1 --fps 30000/1001 -o 111.vvc

[libvvenc @ 000001f236eabd30] libvvenc version: 1.12.1-0f7b70d
[libvvenc @ 000001f236eabd30] vvenc [notice]: Internal format : 1920x1080 29.97 Hz SDR
vvenc [notice]: Threads : 10 (parallel frames: 4)
vvenc [notice]: Rate control : VBR 6 Mbps single-pass
vvenc [notice]: Perceptual optimization : Enabled
vvenc [notice]: Intra period (keyframe) : 256
vvenc [notice]: Decoding refresh type : CRA
vvenc [notice]: Sequence PSNR output : Linear average only
vvenc [notice]: Hexadecimal PSNR output : Disabled
vvenc [notice]: Sequence MSE output : Disabled
vvenc [notice]: Frame MSE output : Disabled
vvenc [notice]: Cabac-zero-word-padding : Enabled
vvenc [notice]: Frame index : all frames
vvenc [notice]: Profile : main_10_444
vvenc [notice]: Level : 4
vvenc [notice]: CU size : 128
vvenc [notice]: Max TB size : 64
vvenc [notice]: Min CB size : 4
vvenc [notice]: Motion search range : 384
vvenc [notice]: QP : 33
vvenc [notice]: Max dQP signaling subdiv : 2
vvenc [notice]: Cb QP Offset (dual tree) : 0 (0)
vvenc [notice]: Cr QP Offset (dual tree) : 0 (0)
vvenc [notice]: GOP size : 32
vvenc [notice]: PicReordering : 1
vvenc [notice]: Input bit depth : (Y:10, C:10)
vvenc [notice]: MSB-extended bit depth : (Y:10, C:10)
vvenc [notice]: Internal bit depth : (Y:10, C:10)
vvenc [notice]: cu_chroma_qp_offset_subdiv : -1
vvenc [notice]: log2_sao_offset_scale_luma : 0
vvenc [notice]: log2_sao_offset_scale_chroma : 0
vvenc [notice]: Cost function : Lossy coding
vvenc [notice]: Film grain analysis : 0

Output #0, vvc, to '112.vvc':
Metadata:
encoder : Lavf61.9.100
Stream #0:0: Video: vvc, 1 reference frame, yuv422p10le(pc, progressive), 1920x1080, q=2-31, 6000 kb/s, 29.97 fps, 29.97 tbn

[Selur]

Hmm, you are using an intermediate file, which I want to avoid.
Also looking at your calls:

ffmpeg_avx2.exe -v verbose -f concat -safe 0 -i mylist_v.txt -f concat -safe 0 -i mylist_a.txt -c:v libvvenc -vb 6000k -c:a aac -ac 2 -ar 48000 -ab 128k -frames:v 100 -movflags faststart -threads 0 -vf scale=1920:1080:in_range=full: out_range=full,format=yuv422p10le 112.vvc
vvencapp_avx2.exe -i 111.y4m -s 1920x1080 -c yuv422 -b 6M -ip 256 --passes 1 --fps 30000/1001 -o 111.vvc

I'm confused.
Why encode the audio when you output 112.vvc?
Where does 111.y4m come from and how can it be input and output?
using:

ffmpeg -y -loglevel fatal -noautorotate -nostdin -threads 8 -i "G:\TestClips&Co\files\test.avi" -map 0:0 -an -sn -pix_fmt yuv422p10le -strict -1 -fps_mode passthrough -f yuv4mpegpipe - | vvencapp --input - --size 640x352 --y4m 1  -c yuv422 --framerate 25 --framescale 1 --internal-bitdepth 10 --bitrate 0 --qp 20 --profile "auto" --level "auto" --tier "main" --preset medium --qpa 1 --sdr sdr_470bg --mtprofile -1 --ifp -1 --intraperiod 250 --output "g:\test_422.vvc"

results in:

Error parsing option "format,c" with argument "yuv422".
Error parsing option "format,c" with argument "yuv422".
error: bitstream file name must be specified (--output=bit.266)

=> sorry, I can't make heads or tails out of your calls.

Cu Selur

[Jamaika]

Originally Posted by Selur

Why encode the audio when you output 112.vvc?

My mistake and rush. You can create MOV in ffmpeg.

Originally Posted by Selur

Where does 111.y4m come from and how can it be input and output?

ffmpeg.exe -y -i film_java_d.avi -f yuv4mpegpipe -pix_fmt yuv422p -frames:v 100 -strict -1 111.y4m

Originally Posted by Selur

using:

ffmpeg -y -loglevel fatal -noautorotate -nostdin -threads 8 -i "G:\TestClips&Co\files\test.avi" -map 0:0 -an -sn -pix_fmt yuv422p10le -strict -1 -fps_mode passthrough -f yuv4mpegpipe - | vvencapp --input - --size 640x352 --y4m 1  -c yuv422 --framerate 25 --framescale 1 --internal-bitdepth 10 --bitrate 0 --qp 20 --profile "auto" --level "auto" --tier "main" --preset medium --qpa 1 --sdr sdr_470bg --mtprofile -1 --ifp -1 --intraperiod 250 --output "g:\test_422.vvc"

results in:

Error parsing option "format,c" with argument "yuv422".
Error parsing option "format,c" with argument "yuv422".
error: bitstream file name must be specified (--output=bit.266)

=> sorry, I can't make heads or tails out of your calls.

And there should be an error. Incorrect input format entered.

[Selur]

using vvencFFapp:

ffmpeg -y -loglevel fatal -noautorotate -nostdin -threads 8 -i "G:\TestClips&Co\files\test.avi" -map 0:0 -an -sn -pix_fmt yuv422p10le -strict -1 -fps_mode passthrough -f yuv4mpegpipe - | vvencFFapp --InputFile - --Size 640x352 --y4m 1 --InputBitDepth 10 --InputChromaFormat 422 --fps 25/1  --TargetBitrate 0 --QP 20 --Profile "auto" --Level "auto" --Tier "main" --preset medium --qpa 1 --sdr sdr_470bg --MTProfile -1 --IFP -1 --IntraPeriod 250 --BitstreamFile "g:\test_422.vvc"

I get:

vvencFFapp: VVenC, the Fraunhofer H.266/VVC Encoder, version 1.12.1-rc1 [Windows][GCC 14.2.0][64 bit][SIMD=AVX2]
Parameter Check Error: Intern chroma format must be either 400, 420
error: input chroma format must be either 400, 420

And there should be an error. Incorrect input format entered.

I do not get it:

  -c,   --format [yuv420]          set input format (yuv420, yuv420_10, yuv420_10_packed, yuv400 (gray), yuv400_10 (gray10)

you used:

vvencapp_avx2.exe -i 111.y4m -s 1920x1080 -c yuv422 -b 6M -ip 256 --passes 1 --fps 30000/1001 -o 111.vvc

So I assumed that you patched that version somehow and yuv422 would be allowed now.
=> I do not get how your '-c yuv422' works for you,..

[Jamaika]

Originally Posted by Selur

So I assumed that you patched that version somehow and yuv422 would be allowed now.
=> I do not get how your '-c yuv422' works for you,..

It's true I didn't change the descriptions. I'm just an amateur.
It should be 'yuv422_10'.
I'm not sure if I wrote it down correctly?

size_t frameSize = (width * height * uiBitsPerPx) >> 3;

  if ( packed && bitdepth == 10 && chromaFormat == VVENC_CHROMA_420 )
  {
    size_t stride = width * 5 / 4;
    size_t lumaSize = stride * height;
    size_t chromaSize = lumaSize >> 2;
    frameSize = lumaSize + chromaSize + chromaSize;
  }

  if ( packed && bitdepth == 10 && chromaFormat == VVENC_CHROMA_422 )
  {
    size_t stride = width / 2;
    size_t lumaSize = stride * height;
    size_t chromaSize = lumaSize >> 2;
    frameSize = lumaSize + chromaSize + chromaSize;
  }

  if ( packed && bitdepth == 10 && chromaFormat == VVENC_CHROMA_444 )
  {
    size_t stride = width;
    size_t lumaSize = stride * height;
    size_t chromaSize = lumaSize >> 2;
    frameSize = lumaSize + chromaSize + chromaSize;
  }

And lastly. If I added it correctly then do we have decoder for yuv 422 and 444.

[Jamaika]

Testing:
Compatibility between different vvc creators in 10bit. There is no player or codec for 8bit.
I haven't tested in two years.

ffmpeg.exe -y -i "Video.mp4" -f yuv4mpegpipe -vf scale=1920:1080:in_range=full:out_range=full,format=yuv420p10le -frames:v 100 -strict -1 111.y4m
ffmpeg.exe -y -i "Video.mp4" -f yuv4mpegpipe -vf scale=1920:1080:in_range=full:out_range=full,format=yuv422p10le -frames:v 100 -strict -1 112.y4m
ffmpeg.exe -y -i "Video.mp4" -f yuv4mpegpipe -vf scale=1920:1080:in_range=full:out_range=full,format=yuv444p10le -frames:v 100 -strict -1 113.y4m

vvencapp_avx2.exe --verbosity 3 -i 111.y4m --y4m -s 1920x1080 -c yuv420_10 --preset medium --fps 30000/1001 -b 6M -ip 256 --passes 1 --threads -1 -o 111.vvc

uvg266_10bit_avx2.exe -i 111.y4m -o 112.vvc --input-file-format y4m --input-res 1920x1080 --input-format P420 --input-bitdepth 10 --preset medium --input-fps 30000/1001 --bitrate 6000000 --vps-period 256 --info --threads auto --range pc

ffmpeg_avx2.exe -y -v info -i "Best of Rallye Rally Crash & Mistakes 2018 by ToutAuCable.mp4" -c:v libvvenc -vb 6000k -r 30000/1001 -frames:v 100 -movflags faststart -threads 0 -vf scale=1920:1080:in_range=full: out_range=full,format=yuv420p10le 113.vvc

EncoderApp_REXT.exe --SummaryVerboseness -c "encoder_randomaccess_vtm.cfg" --InputFile=111.y4m --BitstreamFile=114.vvc --SourceWidth=1920 --SourceHeight=1080 --FrameRate=30000/1001 --InputBitDepth=10 --InternalBitDepth=0 --OutputBitDepth=10 --MSBExtendedBitDepth=10 --InputChromaFormat=420 --ChromaFormatIDC=420 --ConformanceWindowMode=0 --FramesToBeEncoded=100 --MatrixCoefficients=1 --InputColorPrimaries=-1 --LMCSSignalType=0 --Level=5.1 --BDPCM=1 --Tier=main --HashME=1 --IBC=1 --MaxCUWidth=16 --MaxCUHeight=16 --CTUSize=32 --QP=28 --MaxBTLumaISlice=32 --MaxBTChromaISlice=32 --MaxBTNonISlice=32 --MaxTTLumaISlice=32 --MaxTTChromaISlice=32 --MaxTTNonISlice=32 --ColorTransform=0 --VideoFullRange=1 --InputSampleRange=1 --AspectRatioInfoPresent=1 --ChromaLocInfoPresent=1 --Log2MaxTbSize=5 --RCCpbSize=2000 --RateControl=1 --TargetBitrate=6000000 --IntraPeriod=256

EncoderApp_REXT.exe --SummaryVerboseness -c "encoder_randomaccess_vtm.cfg" --InputFile=112.y4m --BitstreamFile=115.vvc --SourceWidth=1920 --SourceHeight=1080 --FrameRate=30000/1001 --InputBitDepth=10 --InternalBitDepth=0 --OutputBitDepth=10 --MSBExtendedBitDepth=10 --InputChromaFormat=422 --ChromaFormatIDC=422 --ConformanceWindowMode=0 --FramesToBeEncoded=100 --MatrixCoefficients=1 --InputColorPrimaries=-1 --LMCSSignalType=0 --Level=5.1 --BDPCM=1 --Tier=main --HashME=1 --IBC=1 --MaxCUWidth=16 --MaxCUHeight=16 --CTUSize=32 --QP=28 --MaxBTLumaISlice=32 --MaxBTChromaISlice=32 --MaxBTNonISlice=32 --MaxTTLumaISlice=32 --MaxTTChromaISlice=32 --MaxTTNonISlice=32 --ColorTransform=0 --VideoFullRange=1 --InputSampleRange=1 --AspectRatioInfoPresent=1 --ChromaLocInfoPresent=1 --Log2MaxTbSize=5 --RCCpbSize=2000 --RateControl=1 --TargetBitrate=6000000 --IntraPeriod=256

EncoderApp_REXT.exe --SummaryVerboseness -c "encoder_randomaccess_vtm.cfg" --InputFile=113.y4m --BitstreamFile=116.vvc --SourceWidth=1920 --SourceHeight=1080 --FrameRate=30000/1001 --InputBitDepth=10 --InternalBitDepth=0 --OutputBitDepth=10 --MSBExtendedBitDepth=10 --InputChromaFormat=444 --ChromaFormatIDC=444 --ConformanceWindowMode=0 --FramesToBeEncoded=100 --MatrixCoefficients=1 --InputColorPrimaries=-1 --LMCSSignalType=0 --Level=5.1 --BDPCM=1 --Tier=main --HashME=1 --IBC=1 --MaxCUWidth=16 --MaxCUHeight=16 --CTUSize=32 --QP=28 --MaxBTLumaISlice=32 --MaxBTChromaISlice=32 --MaxBTNonISlice=32 --MaxTTLumaISlice=32 --MaxTTChromaISlice=32 --MaxTTNonISlice=32 --ColorTransform=0 --VideoFullRange=1 --InputSampleRange=1 --AspectRatioInfoPresent=1 --ChromaLocInfoPresent=1 --Log2MaxTbSize=5 --RCCpbSize=2000 --RateControl=1 --TargetBitrate=6000000 --IntraPeriod=256

Thread options:
VVenc originally has auto thread and this isn't the maximum for a given computer. (multithreading; -1: resolution < 720p: 4, < 5K 2880p: 8, >= 5K 2880p: 12 threads)
UVG266 for max 20 it is 20 but ffmpeg warns that the maximum should be 16.
VTM codec has the most new features added but there isn't thread option.

Playing in ffmpeg vvc:
Color range is always tv.
Frame rate is always 25,000.
UVG266 yuv420 10bit is not standard.

  Stream #0:0: Video: vvc (Main), yuv420p10le(tv), 1920x1080, 25 fps, 29.97 tbr, 1200k tbn
[vvc @ 0000024e446441a0] frame    17, P( 29,  16) failed with -1094995529
[vvc @ 0000024e446441a0] frame    33, P( 29,   5) failed with -1094995529
[vvc @ 0000024e446441a0] frame    34, P( 29,   8) failed with -1094995529
[vvc @ 0000024e446441a0] frame    37, P( 29,   8) failed with -1094995529
[vvc @ 0000024e446441a0] frame    50, P( 12,   7) failed with -1094995529
[vvc @ 0000024e446441a0] frame    55, P( 29,   6) failed with -1094995529
[vvc @ 0000024e446441a0] frame    58, P( 29,   4) failed with -1094995529
[vvc @ 0000024e446441a0] frame    65, P(  5,   7) failed with -1094995529
[vvc @ 0000024e446441a0] frame    66, P( 29,   2) failed with -1094995529
[vvc @ 0000024e446441a0] frame    68, P( 29,   0) failed with -1094995529
[vvc @ 0000024e446441a0] frame    63, P( 29,   5) failed with -1094995529
[vvc @ 0000024e446441a0] frame    74, P( 29,   1) failed with -1094995529
[vvc @ 0000024e446441a0] frame    82, P( 29,   0) failed with -1094995529
[vvc @ 0000024e446441a0] frame    81, P( 29,   1) failed with -1094995529
[vvc @ 0000024e446441a0] frame    91, P( 29,   2) failed with -1094995529
[vvc @ 0000024e446441a0] frame    97, P( 29,   6) failed with -1094995529
-7686143364034.47 M-V:  1.567 fd=   0 aq=    0KB vq=    0KB sq=    0B

ffmpeg with vvenc movie plays too fast.

Video sample:
https://www.sendspace.com/file/h9c4z3

[Jamaika]

ffmpeg.exe -y -i "input.mp4" -f yuv4mpegpipe -vf scale=1920:1080:in_range=full:_out_range=full,format=yuv420p10le -frames:v 100 -strict -1 111.y4m
EncoderApp_VTM_REXT.exe --SummaryVerboseness -c "encoder_randomaccess_vtm.cfg" --InputFile=111.y4m --BitstreamFile=114.vvc --SourceWidth=1920 --SourceHeight=1080 --FrameRate=30000/1001 --InputBitDepth=10 --InternalBitDepth=0 --OutputBitDepth=10 --MSBExtendedBitDepth=10 --InputChromaFormat=420 --ChromaFormatIDC=420 --ConformanceWindowMode=0 --FramesToBeEncoded=100 --MatrixCoefficients=1 --InputColorPrimaries=-1 --LMCSSignalType=0 --Level=5.1 --BDPCM=1 --Tier=main --HashME=1 --IBC=1 --MaxCUWidth=16 --MaxCUHeight=16 --CTUSize=32 --MaxBTLumaISlice=32 --MaxBTChromaISlice=32 --MaxBTNonISlice=32 --MaxTTLumaISlice=32 --MaxTTChromaISlice=32 --MaxTTNonISlice=32 --Log2MaxTbSize=5 --ColorTransform=0 --VideoFullRange=1 --InputSampleRange=1 --AspectRatioInfoPresent=1 --ChromaLocInfoPresent=1 --RCCpbSize=2000 --RateControl=1 --TargetBitrate=6000000 --IntraPeriod=256
VVCSoftware: VTM Encoder Version 23.6-ac3ffaa [Windows][GCC 11.5.0][64 bit] [SIMD=AVX2]
https://www.sendspace.com/file/la2bjr

[benwaggoner]

Originally Posted by Sagittaire

Well if you use x265 with preset placebo or SVT-AV1 with preset 0, you obtain really low speed too. Use the slowest preset for codec is never obligation.

And if x265 --preset placebo isn't enough for you, add
--cu-lossless
--tskip
--subme 7
rc=lookahead==keyint

Which won't make any appreciable quality improvement for most content, but can improve compression efficiency for lossless screen recording, end credits, or other stuff with zero noise and lots of very sharp edges.