- Cin - Cinelerra-GG Mailing Lists

10bit av1 and h264 QSV and VAAPI encoding
by Terje J. Hanssen 19 Nov '24

19 Nov '24

Now when 10bit hevc_qsv and hevc_vaapi accelerated encoding looks to be in box, I wonder if something prohibits us to get accelerated encoding for av1 and h264? According to ffmpeg help the following support should be available: av1_qsv Supported pixel formats: nv12 p010le qsv -profile <int> E..V....... (from 0 to INT_MAX) (default unknown) unknown 0 E..V....... main 1 E..V....... av1_vaapi Supported pixel formats: vaapi -profile <int> E..V....... Set profile (seq_profile) (from -99 to 255) (default -99) main 0 E..V....... high 1 E..V....... professional 2 E..V....... h264_qsv Supported pixel formats: nv12 qsv -profile <int> E..V....... (from 0 to INT_MAX) (default unknown) unknown 0 E..V....... baseline 66 E..V....... main 77 E..V....... high 100 E..V....... h264_vaapi Supported pixel formats: vaapi -profile <int> E..V....... Set profile (profile_idc and constraint_set*_flag) (from -99 to 65535) (default -99) constrained_baseline 578 E..V....... main 77 E..V....... high 100 E..V....... high10 110 E..V.......

3 18

more complete 10bit vaapi encode patches
by Andrew Randrianasulu 15 Nov '24

15 Nov '24

Hopefully you can just put "cin_pix_fmt=p010le" in profile and it will be respected Terje, you only need to apply second patch on top of that you already have.

2 7

another 10 bit encoding patch
by Andrew Randrianasulu 15 Nov '24

15 Nov '24

you need to reset your git tree, something like git reset --hard 2037bf63cdf25462ca606f4a28636e12df77c2a0 before applying this patch then run cin like CIN_10BIT_ENC=p010le bin/cin or replace p010le with whatever equivalents y210le (?) sorry, I just hanged my X session so can't copy/paste exact output, but if you launch cingg with error=debug in bin/ffmpeg/encode.opts it will print list of accepted by vaapi library formats at encoding attempt. I think this is still more like prototype to see if we can get it working, I hope to wire it to existing profile reading mechanism.

2 11

Fwd: New Topic
by Phyllis Smith 14 Nov '24

14 Nov '24

Anyone familiar with this package? Cincity uses CinGG but obviously we all use other packages for specific things? ---------- Forwarded message --------- From: Cinelerra GG - Forum <forum(a)cinelerra-gg.org> Date: Wed, Nov 13, 2024 at 5:41 AM Subject: New Topic To: <phylsmith2017(a)gmail.com> Hello phylsmith2017(a)gmail.com! New topic has been created on your subscribed forum - Miscellaneous <https://www.cinelerra-gg.org/forum/forum-miscellaneous/>. *Help on slowmoVideo* I need to slow down a clip which has been recorded at 30 FPS. slowmoVideo is very interesting as it uses the optical flow technology for interpolating frames like respeedr does. And it uses FFMpeg. It can produce super fluid video. It can even be enhanced by a module to use the GPU efficiently. Unfortunately I have problems and at Github no one is answering to this technical issue (maybe because the people spammed around in the past). Does anyone of you have the possibility to take a look at it? My debian package installs perfectly and starts without issues. I can open the clip but on render it crashes. Using the command line I get this error: Have OpenCL: true useOpenCL: true true 1 1 terminate called after throwing an instance of 'cv::Exception' what(): OpenCV(4.6.0) ./modules/video/src/optflowgf.cpp:1120: error: (-215:Assertion failed) _flow0.size() == prev0.size() && _flow0.channels() == 2 && _flow0.depth() == CV_32F in fun... ------------------------------ If you want to unsubscribe from this forum please use the link below.

3 2

10bit VAAPI encoding preset
by Terje J. Hanssen 13 Nov '24

13 Nov '24

After finishing the three hevc_qsv presets for _8b420, _10b420 and _10b422 mp4 https://lists.cinelerra-gg.org/pipermail/cin/2024-November/009039.html I thought to do the corresponding for hevc_vaapi However I am struggling to get 10bit 422 output rendered from Cingg, which so far falls back to 10bit 420 chroma subsampling fprobe -hide_banner hd01_hevc_vaapi_10b422.mp4 Duration: 00:01:11.20, start: 0.000000, bitrate: 11082 kb/s Stream #0:0[0x1](und): Video: hevc (Main 10) (hev1 / 0x31766568), yuv420p10le(tv, bt709/unknown/unknown, top coded first (swapped)), 1920x1080 [SAR 1:1 DAR 16:9], 11080 kb/s, 25 fps, 25 tbr, 12800 tbn (default) In a previous thread I proved that ffmpeg hevc_vaapi encoding has this 10bit-422 capability 3) 10-bit yuv422p10le hevc_vaapi encoding using -vf 'format=y210,hwupload' https://lists.cinelerra-gg.org/pipermail/cin/2024-November/008982.html And yes, there was also issues before hevc_qsv_10b422.mp4 combinations Pixel format/profile finally succeeded. But here are some differences, yet. While hevc_qsv provided a Pixels drop down menu with pixel formats, hevc_vaapi just has a single "vaapi" pixel format. Any idea why, and what can fulfill the vaapi 10bit 422 encoding here?

3 8

10bit VAAPI encoding test thread continued
by Terje J. Hanssen 13 Nov '24

13 Nov '24

Continued and extracted from previous threads_ https://lists.cinelerra-gg.org/pipermail/cin/2024-October/008967.html Den 27.10.2024 21:11, skrev Terje J. Hanssen: > ........snip > > For me it looks like nothing has changed: > > localhost:/Cin # git am 0001-Experimental-try-10bit-vaapi-encoding.patch > fatal: previous rebase directory .git/rebase-apply still exists but > mbox given. > > > remove it again? > > > By the way,testing > > bin/cin > compression: hevc_vaapi.mp4 > Pixels: vaapi (no other options) > cin_hw_dev=vaapi > > Renders > > Color space : YUV > Chroma subsampling : 4:2:0 > Bit depth : 8 bits > > > The same with > CIN_10BIT_ENC=1 bin/cin > > > > yeah, it does not exist in normal cin, only after you add patch. > > > may be just take a break from all this, I do not think we will solve > this problem in few remaining days until end of October. So ... do > not rush > > Ok for me. > Den 04.11.2024 21:49, skrev Andrew Randrianasulu: > > Terje, I think if you are ok with this idea we can return to testing > 10bit-vaapi patch ... > Hopefully in its final form it will just allow same format= string in > encoding profiles as supported by per-file decoding opts files now ,,, > I'm not sure I understood the last line yet, but I should be ready to continue vaapi-testing to-morrow. ==================== Yeah, here we are. The problem as I understood lastly, was to apply the 10bit-vaapi-encoding.patch?

3 35

FFmpeg supported QSV pixel formats and profiles
by Terje J. Hanssen 11 Nov '24

11 Nov '24

An overview over the system FFmpeg 7.1 supported QSV pixel formats and profiles: ffmpeg -hide_banner -encoders | grep qsv V..... av1_qsv AV1 (Intel Quick Sync Video acceleration) (codec av1) V..... h264_qsv H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (Intel Quick Sync Video acceleration) (codec h264) V..... hevc_qsv HEVC (Intel Quick Sync Video acceleration) (codec hevc) V..... mjpeg_qsv MJPEG (Intel Quick Sync Video acceleration) (codec mjpeg) V..... mpeg2_qsv MPEG-2 video (Intel Quick Sync Video acceleration) (codec mpeg2video) V..... vp9_qsv VP9 video (Intel Quick Sync Video acceleration) (codec vp9) ffmpeg -hide_banner -h encoder=av1_qsv | egrep 'Supported|profile|main|high|professional' Supported hardware devices: qsv qsv qsv Supported pixel formats: nv12 p010le qsv -profile <int> E..V....... (from 0 to INT_MAX) (default unknown) main 1 E..V....... ffmpeg -hide_banner -h encoder=hevc_qsv | egrep 'Supported|profile|main|high|professional' Supported hardware devices: qsv qsv qsv Supported pixel formats: nv12 p010le p012le yuyv422 y210le qsv bgra x2rgb10le vuyx xv30le -profile <int> E..V....... (from 0 to INT_MAX) (default unknown) main 1 E..V....... main10 2 E..V....... mainsp 3 E..V....... -tier <int> E..V....... Set the encoding tier (only level >= 4 can support high tier) (from 0 to 256) (default high) main 0 E..V....... high 256 E..V....... ffmpeg -hide_banner -h encoder=h264_qsv | egrep 'Supported|profile|main|high|professional' Supported hardware devices: qsv qsv qsv Supported pixel formats: nv12 qsv -profile <int> E..V....... (from 0 to INT_MAX) (default unknown) main 77 E..V....... high 100 E..V.......

2 7

Where are Cingg's ffmpeg compression files located?
by Terje J. Hanssen 08 Nov '24

08 Nov '24

During render testing to and fro, I have had a suspect that now and then the content of ffmpeg compression type has changed or could not be picked up again. But I have thought it was me that had confused or forgot something. Or may this occur by other reasons? Seemingly this happened recently during render switching between hevc_vaapi,mp4 and hevc_qsv.mp4 . hevc_qsv.mp4 has several more pixel options than hevc_vaapi.mp4 which has only the single "vaapi" option. And in one occasion I got more pixel options for hevc_vaapi.mp4(?) - or was it still on hevc_qsv,mp4? So to find out, I have tried to edit the hevc_vaapi.mp4 file, but have not got the edited content loaded in the video wrench compression window. So I wonder where is the preset file used by Cingg located? I have # ls -l /Cin lrwxrwxrwx 1 root root 29 Oct 22 20:23 /Cin -> /home/cinelerra/cinelerra-5.1 and the hevc_vaapi.mp4 preset file with the same content located in both /Cin/bin/ffmpeg/video /Cin/ffmpeg/video But the edited content isn't visible in Cingg?

2 7

Enable Cingg single-user build to use system libSvtAv1Enc upgraded to v. 2.3.0
by Terje J. Hanssen 07 Nov '24

07 Nov '24

I upgraded my system Slowroll to S | Name | Type | Version | Arch | Repository ---+---------------------+---------+--------------------------------+--------+---------------------- i+ | libSvtAv1Enc2 | package | 2.3.0-1.5 | x86_64 | obs:multimedia:libs i+ | SVT-AV1 | package | 2.3.0-1.5 | x86_64 | obs:multimedia:libs Version test: # SvtAv1EncApp Svt[info]: ------------------------------------------- Svt[info]: SVT [version]: SVT-AV1 Encoder Lib v2.3.0 Svt[info]: SVT [build] : GCC 14.2.1 20241007 [revision 4af44f2cf7d281f3e4f3957efce10e8b2ccb2ad3] 64 bit Svt[info]: LIB Build date: Oct 29 2024 22:11:44 Svt[info]: ------------------------------------------- File list for rpm -ql libSvtAv1Enc2 /usr/lib64/libSvtAv1Enc.so.2 /usr/lib64/libSvtAv1Enc.so.2.3.0 /usr/share/licenses/libSvtAv1Enc2 /usr/share/licenses/libSvtAv1Enc2/LICENSE.md /usr/share/licenses/libSvtAv1Enc2/PATENTS.md I tested first that SVT-AV1 v. 2.3.0 encoding worked OK with my existing system ffmpeg v. 7.1 I'm not sure I ever have tested av1_svt rendering with my single-user build Cingg before now. But when I tried now to render the same hdv file as with ffmpeg, Cingg failed with int FFMPEG::open_encoder(const char*, const char*): cant find codec libsvtav1:/Videoklipp/AV1/hdv09_04_m2t_svt-av1_230.webm Shouldn't Cingg be able to work with the system upgraded libSvtAv1Enc2 v. 2-3-0 automatically as well as ffmpeg does? Or if necessary via a simple "update", without complete rebuild? ------------- I had a look at my last rebuild of Cingg for using ffmpeg 7.1 : Does "disabled libsvtav1" say something? # ./configure --with-single-user --disable-static-build --without-thirdparty --without-libdpx configure: loading site script /usr/share/site/x86_64-pc-linux-gnu ..........snip checking for /usr/local/cuda/include/cuda.h... no checking cuda sdk... no checking for dlopen in -ldl... yes checking for numa_alloc in -lnuma... yes checking for openexr available... yes Reason Package ------ ------- disabled encore disabled audiofile disabled esound disabled libsvtav1 ...snip...

2 4

New SVT-AV1 release: v2.3.0 from AOM
by Terje J. Hanssen 05 Nov '24

05 Nov '24

As seen on reddit, the Alliance for Open Media - AOM has released SVT-AV1 v. 2.3.0 The Scalable Video Technology for AV1 (SVT-AV1 Encoder) is an AV1-compliant software encoder library. * NEW FAST DECODE MODE * New fast-decode (2) to allow for an average AV1 software cycle reduction of 25-50% vs fast-decode 0 with a 1-3% BD-Rate loss across the presets * Improved fast-decode (1) option to increase its AV1 software cycle reduction by ~10% while maintaining the same quality levels * Improved --lp settings for high resolutions, with CRF gaining a ~4% improvement in speed and VBR gaining ~15% * Further Arm-based optimizations improving the efficiency of previously written Arm-neon implementations by an average of 30% * Address speed regressions for high resolutions first pass encode by tuning the threading parameters, with 1080p showing the biggest gains * Enabled AVX512 by default in cmake allowing for ~2-4% speedup * ...and more https://gitlab.com/AOMediaCodec/SVT-AV1/-/releases/v2.3.0#230---2024-10-28 https://gitlab.com/AOMediaCodec/SVT-AV1/-/blob/master/README.md?ref_type=he…

3 5