vpp-resize mfx_ai_superres

[hmartin417]

I noticed that this was added in a past release as a beta sort of option. I am unable to get this to work. I have even tried to uninstall the MSDK and installed oneVPL that supports this but still get an error trying to use this. Is this something that you can look into to, to have this option work? I'm currently on Server 2019 with 12th gen I9 and A750 GPU.

[rigaya]

mfx_ai_superres already works but it seems to have following limitations.

• Requires Arc GPU.
• upscaling only.
• output height must be 1080 or below.
• 8bit only (does not work for 10bit)

I'm not sure about OS support, at least it works on Win11 with the limitation written above.

[hmartin417]

So i would have to upscale first alone and then use qsvenc a 2nd time to do a regular encode to av1 or hevc?

…

On Sat, Nov 23, 2024, 6:54 PM rigaya ***@***.***> wrote: mfx_ai_superres already works but it seems to have following limitations. - Requires Arc GPU. - upscaling only. - output height must be 1080 or below. - 8bit only (does not work for 10bit) I'm not sure about OS support, at least it works on Win11 with the limitation written above. — Reply to this email directly, view it on GitHub <#224 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AMZBZIPCRFAKEBPW4LTBX7L2CEIUHAVCNFSM6AAAAABSLMIMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJVGY4TSNZUGU> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[rigaya]

You can still encode to av1 or hevc if it is within the limitation I've written above, 8bit and output height is 1080 or below.

If the resolution or bit depth exceeds the limitation, you cannot use mfx_ai_superres due to driver (or device) limitations, so you will need to use other resize algorithm or think of other ways.

[hmartin417]

Could you provide a cli example so i can compare it against my current cli command?

…

On Sat, Nov 23, 2024, 7:09 PM Harry Martin ***@***.***> wrote: So i would have to upscale first alone and then use qsvenc a 2nd time to do a regular encode to av1 or hevc? On Sat, Nov 23, 2024, 6:54 PM rigaya ***@***.***> wrote: > mfx_ai_superres already works but it seems to have following limitations. > > - Requires Arc GPU. > - upscaling only. > - output height must be 1080 or below. > - 8bit only (does not work for 10bit) > > I'm not sure about OS support, at least it works on Win11 with the > limitation written above. > > — > Reply to this email directly, view it on GitHub > <#224 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AMZBZIPCRFAKEBPW4LTBX7L2CEIUHAVCNFSM6AAAAABSLMIMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJVGY4TSNZUGU> > . > You are receiving this because you authored the thread.Message ID: > ***@***.***> >

[rigaya]

Sure. Following is example of 1280x720 → 1920x1080 upscale, encoding H.264 to AV1.

x64\QSVEncC64.exe -i Y:\Encoders\sakura_op.mp4 -o F:\temp\test.mp4 -c av1 --output-res 1920x1080 --vpp-resize mfx_a
i_superres

QSVEncC (x64) 7.74 (r3509) by rigaya, Nov 22 2024 11:34:10 (VC 1942/Win)
OS             Windows 11 x64 (26100) [UTF-8]
CPU Info       12th Gen Intel Core i9-12900K [5.02GHz] (8P+8E,16C/24T) <DG2>
GPU Info       Intel Arc A380 Graphics (128EU) 300-2450MHz (32.0.101.6127)
Media SDK      QuickSyncVideo (hardware encoder) FF, 1st GPU(d), API v2.13
Async Depth    3 frames
Hyper Mode     off
Buffer Memory  d3d11, 41 work buffer
Input Info     avqsv: H.264/AVC, 1280x720, 30/1 fps
VPP            AI SuperRes, mode 1
               Resize 1280x720 -> 1920x1080
AVSync         auto
Output         AV1(yuv420) main @ Level 4
               1920x1080p 1:1 30.000fps (30/1fps)
               avwriter: av1 => mp4
Target usage   4 - balanced
Encode Mode    ICQ (Intelligent Const. Quality)
ICQ Quality    23
QP Limit       min: 1, max: 255
Ref frames     4 frames
GopRefDist     8, B-pyramid: on
Max GOP Length 300 frames

encoded 3501 frames, 61.17 fps, 3158.23 kbps, 43.94 MB
encode time 0:00:57, CPU: 2.0%, GPU: 78.3%, VD: 11.6%
frame type IDR   12
frame type I     12,  total size   0.83 MB
frame type P   3489,  total size  43.11 MB

[hmartin417]

Thanks! I'm guessing that i cant copy audio tracks at the same time?

[rigaya]

You shall be able to copy all audio tracks by --audio-copy.

[hmartin417]

Im still unable to get this to work still. I must be missing something somewhere with my sdk

…

On Sun, Nov 24, 2024, 4:11 AM rigaya ***@***.***> wrote: You shall be able to copy all audio tracks by --audio-copy. — Reply to this email directly, view it on GitHub <#224 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AMZBZIOJXM4B3YL277I6CF32CGJ3VAVCNFSM6AAAAABSLMIMTKVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDIOJVHA4DGNJUGY> . You are receiving this because you authored the thread.Message ID: ***@***.***>