[FFmpeg-user] h264_qsv encoder doesn't support changing input dimensions

[Stepan Salenikovich]

Hi, I have a cmd like this to convert mjpeg into h264 using the qsv hw
encoder on windows:

ffmpeg.exe -loglevel debug -use_wallclock_as_timestamps 1 -f mjpeg -hwaccel
qsv -c:v mjpeg_qsv -i - -movflags empty_moov -frag_duration 100000 -nostats
-metadata rotate=0 -vf format=qsv -c:v h264_qsv -flush_packets 1 file.mp4

This works until the size of the incoming jpeg changes and there is an
error and ffmpeg exits (log follows at the end of the message).
I've tried adding a scaling filter so that the dimensions of the output
stay the same, eg:

ffmpeg.exe -loglevel debug -use_wallclock_as_timestamps 1 -f mjpeg -hwaccel
qsv -c:v mjpeg_qsv -i - -movflags empty_moov -frag_duration 100000 -nostats
-metadata rotate=0 -vf vpp_qsv='width=iw+mod(iw,2):height=ih+mod(ih,2)'
-c:v h264_qsv -flush_packets 1 file.mp4

But this doesn't seem to change much. I get a very similar debug log and
error.

Using libx264 I don't have such an issue. So is this a limitation/bug in
the qsv_h264 encoder? Or am I missing something in my command?

Input #0, mjpeg, from 'pipe:':
  Duration: N/A, start: 1629142556.362805, bitrate: N/A
  Stream #0:0, 1, 1/1200000: Video: mjpeg (Baseline), 1 reference frame,
yuvj420p(pc, bt470bg/unknown/unknown, center), 1925x939 [SAR 1:1 DAR
1925:939], 0/1, 25 tbr, 1200k tbn$
 25 tbc
Successfully opened the file.
Parsing a group of options: output url
/tmp/tmp4CrSJa-video/266d372a-fec9-11eb-b8b7-00155d4b0807.mp4.
Applying option metadata (add metadata) with argument rotate=0.
Applying option vf (set video filters) with argument format=qsv.
Applying option c:v (codec name) with argument h264_qsv.
Successfully parsed a group of options.
Opening an output file:
/tmp/tmp4CrSJa-video/266d372a-fec9-11eb-b8b7-00155d4b0807.mp4.
[file @ 0000025f7a384680] Setting default whitelist 'file,crypto,data'
Successfully opened the file.
Stream mapping:
  Stream #0:0 -> #0:0 (mjpeg (mjpeg_qsv) -> h264 (h264_qsv))
cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if
it occurs once at the start per stream)
[AVHWDeviceContext @ 0000025f7a386300] Using D3D9Ex device.
[AVHWDeviceContext @ 0000025f7a386200] Initialize MFX session: API version
is 1.34, implementation version is 1.32
[mjpeg_qsv @ 0000025f7a3810c0] Format qsv chosen by get_format().
[mjpeg_qsv @ 0000025f7a3810c0] Decoder: output is video memory surface
[mjpeg_qsv @ 0000025f7a3810c0] Format qsv chosen by get_format().
[mjpeg_qsv @ 0000025f7a3810c0] Decoder: output is video memory surface
cur_dts is invalid st:0 (0) [init:0 i_done:0 finish:0] (this is harmless if
it occurs once at the start per stream)
    Last message repeated 2 times
Frame parameters mismatch context 1926,940,116 != 1926,940,23
detected 8 logical cores
[Parsed_format_0 @ 0000025f7a3f9b00] Setting 'pix_fmts' to value 'qsv'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] Setting 'video_size' to
value '1926x940'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] Setting 'pix_fmt' to
value '116'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] Setting 'time_base' to
value '1/1200000'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] Setting 'pixel_aspect'
to value '1/1'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] Setting 'frame_rate' to
value '25/1'
[graph 0 input from stream 0:0 @ 0000025f7a3f9c80] w:1926 h:940 pixfmt:qsv
tb:1/1200000 fr:25/1 sar:1/1
[format @ 0000025f7a3dad80] Setting 'pix_fmts' to value 'nv12|p010le|qsv'
[AVFilterGraph @ 0000025f7a3e9180] query_formats: 4 queried, 3 merged, 0
already done, 0 delayed
[h264_qsv @ 0000025f7a383340] Using input frames context (format qsv) with
h264_qsv encoder.
[h264_qsv @ 0000025f7a383340] Encoder: input is video memory surface
[h264_qsv @ 0000025f7a383340] Using the average variable bitrate (AVBR)
ratecontrol method
[h264_qsv @ 0000025f7a383340] profile: high; level: 40
[h264_qsv @ 0000025f7a383340] GopPicSize: 250; GopRefDist: 4; GopOptFlag:
closed ; IdrInterval: 0
[h264_qsv @ 0000025f7a383340] TargetUsage: 4; RateControlMethod: AVBR
[h264_qsv @ 0000025f7a383340] TargetKbps: 1000; Accuracy: 1; Convergence:
1; BRCParamMultiplier: 1
[h264_qsv @ 0000025f7a383340] NumSlice: 1; NumRefFrame: 3
[h264_qsv @ 0000025f7a383340] RateDistortionOpt: OFF
[h264_qsv @ 0000025f7a383340] RecoveryPointSEI: OFF IntRefType: 0;
IntRefCycleSize: 0; IntRefQPDelta: 0
[h264_qsv @ 0000025f7a383340] MaxFrameSize: 702720; MaxSliceSize: 0;
[h264_qsv @ 0000025f7a383340] BitrateLimit: ON; MBBRC: OFF; ExtBRC: OFF
[h264_qsv @ 0000025f7a383340] Trellis: auto
[h264_qsv @ 0000025f7a383340] VDENC: OFF
[h264_qsv @ 0000025f7a383340] RepeatPPS: OFF; NumMbPerSlice: 0;
LookAheadDS: 2x
[h264_qsv @ 0000025f7a383340] AdaptiveI: OFF; AdaptiveB: OFF; BRefType:
pyramid
[h264_qsv @ 0000025f7a383340] MinQPI: 0; MaxQPI: 0; MinQPP: 0; MaxQPP: 0;
MinQPB: 0; MaxQPB: 0
[h264_qsv @ 0000025f7a383340] Entropy coding: CABAC; MaxDecFrameBuffering: 3
[h264_qsv @ 0000025f7a383340] NalHrdConformance: OFF; SingleSeiNalUnit: ON;
VuiVclHrdParameters: OFF VuiNalHrdParameters: OFF
[h264_qsv @ 0000025f7a383340] FrameRateExtD: 1; FrameRateExtN: 25
[mp4 @ 0000025f7a3816c0] Empty MOOV enabled; disabling automatic bitstream
filtering
Output #0, mp4, to
'/tmp/tmp4CrSJa-video/266d372a-fec9-11eb-b8b7-00155d4b0807.mp4':
  Metadata:
    rotate          : 0
    encoder         : Lavf58.76.100
  Stream #0:0, 0, 1/12800: Video: h264, 1 reference frame (avc1 /
0x31637661), qsv(pc, bt470bg/unknown/unknown, progressive, center),
1926x940 (0x0) [SAR 1:1 DAR 963:470], 0/1,
 q=2-31, 1000 kb/s, 25 fps, 12800 tbn
    Metadata:
      encoder         : Lavc58.134.100 h264_qsv
    Side data:
      cpb: bitrate max/min/avg: 0/0/1000000 buffer size: 0 vbv_delay: N/A
Clipping frame in rate conversion by 0.000008
cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if
it occurs once at the start per stream)
Frame parameters mismatch context 1926,940,116 != 1926,940,23
Clipping frame in rate conversion by 0.033638
cur_dts is invalid st:0 (0) [init:1 i_done:0 finish:0] (this is harmless if
it occurs once at the start per stream)
Frame parameters mismatch context 1926,940,116 != 1926,940,23
*** 7 dup!

...

[mjpeg_qsv @ 0000025f7a3810c0] Format qsv chosen by get_format().
[mjpeg_qsv @ 0000025f7a3810c0] Decoder: output is video memory surface
Frame parameters mismatch context 1926,940,116 != 1538,750,23
*** 3 dup!
[mjpeg_qsv @ 0000025f7a3810c0] Format qsv chosen by get_format().
[mjpeg_qsv @ 0000025f7a3810c0] Decoder: output is video memory surface
Frame parameters mismatch context 1926,940,116 != 1152,748,23
*** 6 dup!
Frame parameters mismatch context 1926,940,116 != 1152,748,23
*** 3 dup!
Frame parameters mismatch context 1538,750,116 != 1152,748,23
[Parsed_format_0 @ 0000025f7a38de00] Setting 'pix_fmts' to value 'qsv'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] Setting 'video_size' to
value '1538x750'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] Setting 'pix_fmt' to
value '116'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] Setting 'time_base' to
value '1/1200000'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] Setting 'pixel_aspect'
to value '1/1'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] Setting 'frame_rate' to
value '25/1'
[graph 0 input from stream 0:0 @ 0000025f7a38d900] w:1538 h:750 pixfmt:qsv
tb:1/1200000 fr:25/1 sar:1/1
[scaler_out_0_0 @ 0000025f7a38df00] Setting 'w' to value '1926'
[scaler_out_0_0 @ 0000025f7a38df00] Setting 'h' to value '940'
[scaler_out_0_0 @ 0000025f7a38df00] Setting 'flags' to value 'bicubic'
[scaler_out_0_0 @ 0000025f7a38df00] w:1926 h:940 flags:'bicubic' interl:0
[format @ 0000025f7a38c900] Setting 'pix_fmts' to value 'qsv'
[auto_scaler_0 @ 0000025f7a38d500] Setting 'flags' to value 'bicubic'
[auto_scaler_0 @ 0000025f7a38d500] w:iw h:ih flags:'bicubic' interl:0
[scaler_out_0_0 @ 0000025f7a38df00] auto-inserting filter 'auto_scaler_0'
between the filter 'Parsed_format_0' and the filter 'scaler_out_0_0'
Impossible to convert between the formats supported by the filter
'Parsed_format_0' and the filter 'auto_scaler_0'
Error reinitializing filters!
Failed to inject frame into filter network: Function not implemented
Error while processing the decoded data for stream #0:0
[AVIOContext @ 0000025f7a3850c0] Statistics: 0 seeks, 161 writeouts
[AVIOContext @ 0000025f7a37ff00] Statistics: 1041917 bytes read, 0 seeks
Conversion failed!