Audio speed - nova-video-player/aos-AVP GitHub Wiki
Design:
timestamps derivation are not changed by audio_speed
when audio_speed>1, parser is outputting audio/video quicker with smaller time units
this means realtime=stream->video_time*audio_speed to scale it back to real value, and conversely stream->video_time=real_time/audio_speed
Notations:
rt := real video|audio time
as := audio speed (2x -> 2.0f), 0.25<=as<=2.00 (audio_interface_get_audio_speed())
mspf := milli seconds per frame (s->video->msPerFrame)
QUESTIONS
should video->bytesPerSec be scaled too?
From audio_interface_audiotrack_java
sets the audio speed and configures PlayParams of AudioTrack to set/change output audio rate without pitch
From stream_parser:
_get_XXX_cdata: chunck data cdata->time is rt/as
_get_XXX_time: time is rt
GET_XXX_TS: timestamps are rt
_seek: need to seek to _get_XXX_time/as i.e. rt/as
_calc_rate: TODO why should it be changed??? called by stream_video:_parser_thread and used to modify vtime_parsed/vcurrent_rate and atime_parsed/acurrent_rateTODO where is it used and should it be scaled?
From stream
stream_set_av_speed is called for a change of as and sets AudioTrack audio speed via audio_interface_change_audio_speed and stream_seek_time to current stream rt i.e. stream_get_current_time with STREAM_SEEK_BACKWARD
From stream_video
_output_frame_no_resize: s->video_time and s->sink_ref_time are rt/as and increments are (mspf/as)
stream_get_time_default: returns rt
QUESTION in _real_time there is s->speed concept that could be used instead with mult & div?
QUESTION_put_frame_in_sink calls _real_time, should it be updated to take as?
TOCHECK_video_decode this is where frame_time is set
From stream_audio
TOCHECK no modification there
From stream_subtitles
_sub_decode for internal subs operates on cdata->time thus requires s->video_time and not rt but external subs operate on rt and thus requires as*s->video_time
From stream_sync
_output_frame_no_resizes->sink_ref_time increments/decrements are mspf/as