[FFmpeg-devel] [PATCH] avformat/dv: always set audio packet duration

[Dave Rice]

> On Sep 6, 2021, at 2:28 PM, Andreas Rheinhardt <andreas.rheinhardt at outlook.com> wrote:
> 
> Paul B Mahol:
>> If audio packet is present in DV stream it have duration of 1 in DV timebase units.
>> 
>> Signed-off-by: Paul B Mahol <onemda at gmail.com>
>> ---
>> libavformat/dv.c | 4 ++++
>> 1 file changed, 4 insertions(+)
>> 
>> diff --git a/libavformat/dv.c b/libavformat/dv.c
>> index d7909683c3..b2b74162df 100644
>> --- a/libavformat/dv.c
>> +++ b/libavformat/dv.c
>> @@ -48,6 +48,7 @@ struct DVPacket {
>>     int      stream_index;
>>     int      flags;
>>     int64_t  pos;
>> +    int64_t  duration;
>> };
>> 
>> struct DVDemuxContext {
>> @@ -276,6 +277,7 @@ static int dv_extract_audio_info(DVDemuxContext *c, const uint8_t *frame)
>>             c->audio_pkt[i].stream_index = c->ast[i]->index;
>>             c->audio_pkt[i].flags       |= AV_PKT_FLAG_KEY;
>>             c->audio_pkt[i].pts          = AV_NOPTS_VALUE;
>> +            c->audio_pkt[i].duration     = 0;
>>             c->audio_pkt[i].pos          = -1;
>>         }
>>         c->ast[i]->codecpar->sample_rate    = dv_audio_frequency[freq];
>> @@ -374,6 +376,7 @@ int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
>>             pkt->stream_index = c->audio_pkt[i].stream_index;
>>             pkt->flags        = c->audio_pkt[i].flags;
>>             pkt->pts          = c->audio_pkt[i].pts;
>> +            pkt->duration     = c->audio_pkt[i].duration;
>>             pkt->pos          = c->audio_pkt[i].pos;
>> 
>>             c->audio_pkt[i].size = 0;
>> @@ -404,6 +407,7 @@ int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
>>         c->audio_pkt[i].pos  = pos;
>>         c->audio_pkt[i].size = size;
>>         c->audio_pkt[i].pts  = (c->sys->height == 720) ? (c->frames & ~1) : c->frames;
>> +        c->audio_pkt[i].duration = 1;
>>         ppcm[i] = c->audio_buf[i];
>>     }
>>     if (c->ach)
>> 
> Sure about that? According to the code, the packets contain slightly
> different amounts of samples; see dv_extract_audio_info(), in particular
> lines 242 and 289.
> (It seems to me that one audio packet is supposed to contain the audio
> for one video frame, yet the standard sample rates and standard video
> framerates (in particular NTSC) are incompatible, so a slight variation
> in the number of samples per frame is used.)
> IMO one should set the audio timebase to be the inverse of the
> samplerate and set the duration based upon the number of samples. But I
> have zero experience with dv.

Yes the audio duration is variable in NTSC DV, see page 17 of http://web.archive.org/web/20060927044735/http://www.smpte.org/smpte_store/standards/pdf/s314m.pdf. Each frame has either 1600 or 1602 samples, so while the video is 1001/30000 the audio is either 1600/48000 or 1602/48000. So video at ~0.03336666667 and audio at either ~0.03333333333 or 0.033375.

I’ve tested this patch and it does correct some errors. For instance, before this patch frame 3 of the output only offers half of the needed duration, so the resulting audio track is (1001/30000)/2 shorter than the video. But after the patch, the video and audio outputs have the same length.

ffmpeg -i https://archive.org/download/test_a_202108/TESTTHIS.dv -filter_complex "[0:a:0]aresample=async=1:min_hard_comp=0.01[aud]" -c:v:0 copy -map 0:v:0 -c:a pcm_s16le -map "[aud]” output.mov

Dave Rice