#9184: pngdec output frame that has double image
-------------------------------------+-------------------------------------
Reporter: akoryagin | Owner: (none)
Type: defect | Status: open
Priority: important | Component: avfilter
Version: unspecified | Resolution:
Keywords: artefact on | Blocked By:
the video |
Blocking: | Reproduced by developer: 0
Analyzed by developer: 0 |
-------------------------------------+-------------------------------------
Comment (by wenbin,chen):
Hi Pljushevij
CPU High because it use cpu to decode. Here is the command using qsv to
decode.(I add "-c:v hevc_qsv" option)
{{{
ffmpeg -init_hw_device qsv=hw:/dev/dri/renderD128 -filter_hw_device hw -re
-f concat -safe 0 -c:v hevc_qsv -i "files.txt" -loop 1 -i "logo.png"
-filter_complex
"[1]lut=a=val*0.5,hwupload=extra_hw_frames=10,format=qsv[a];[0:v]hwupload=extra_hw_frames=10,scale_qsv=1920:-1[bg];[bg][a]overlay_qsv=x=(W-w-1):y=(H-h-1):shortest=1,hwdownload,format=nv12,pad=1920:1080
:(ow-iw)/2:(oh-ih)/2" -map 0:v:0 -map 0:a:0 -c:v hevc_qsv -preset veryfast
-b:v 7M -maxrate 7M -c:a eac3 -sn -f hls -hls_list_size 5 -hls_flags
delete_segments -y stream.m3u8
}}}
There are upload and download parts which still reduces the efficiency. If
we remove pad filter we can remove hwdownload filter too. This makes
encoder to encode directly in device memory.
As for upload part, we need to submit patch to improve it. overlay_qsv
need two input frames to be in the same device context, but ffmpeg-qsv
cannot specify device context for each decoder, so I use hwupload to make
sure they are in the same device context.
--
Ticket URL: <https://trac.ffmpeg.org/ticket/9184#comment:29>
FFmpeg <https://ffmpeg.org>
FFmpeg issue tracker_______________________________________________
FFmpeg-trac mailing list
[email protected]
https://ffmpeg.org/mailman/listinfo/ffmpeg-trac
To unsubscribe, visit link above, or email
[email protected] with subject "unsubscribe".
