The output pts of audio filter 'atempo' seems to be wrong.

[elite_jwp]

At libavfilter/af_atempo.c:1061, the code is as below:

    atempo->dst_buffer->pts = atempo->start_pts +
        av_rescale_q(atempo->nsamples_out,
                     (AVRational){ 1, outlink->sample_rate },
                     outlink->time_base);

According to tempo factor, the number of output samples may be different with that of input samples, but they are equal in time duration.
Generally, inlink->sample_rate is same with outlink->sample_rate. So it seems that the duration of nsamples_out can't be computed like above code. It should take count of tempo factor, and can be fixed like below?

    atempo->dst_buffer->pts = atempo->start_pts +
        av_rescale_q(atempo->nsamples_out,
                     (AVRational){ 1, (int)(outlink->sample_rate/atempo->tempo) },
                     outlink->time_base);

[Elon Musk]

No, that is logic failure. Code is correct.
Atempo changes duration of audio so pts must change too.

[elite_jwp]

Ok. Thanks.
But here exists a case:
I want to change the playback speed in ffplay.exe. So i used the atempo filter to do it.
With 'AV_SYNC_AUDIO_MASTER' mode in ffplay, video is synchronized to audio.
If only changing the audio tempo, it will cause the pts diffrence too big and nonsynchronized between video and audio.
Should i also change the video speed for that purpose?

[Elon Musk]

Yes, you need to change video pts. Note that bug report site is not for general help around ffmpeg. For more info ask elsewhere.