Paint-model subtitles in Shaka Player
Switch to dash.js →Compare the experimental stpc/wvtc tracks with ordinary
stpp/wvtt on the same low-latency livesim2 stream: bytes on the wire, and
how many TTML documents the player has to parse. The 4CCs stpc, wvtc,
ttmn, ttmb and vttn are placeholders, not registered with MP4RA.
The design is described in the paint-model-subtitles README.
Active cue: –
Player work, selected track
Counting since the track was selected.
Parse counts and parse time come from the experimental
shaka.text.Mp4TtmlParser.getParseStats() and
shaka.text.Mp4VttParser.getParseStats() of this build, which count every document
handed to the TTML parser and every WebVTT cue box read. Samples restated from a
ttmn/vttn are never parsed.
Payload bytes on the wire, same segment, all four tracks
| Track | Segment | CMAF bytes | CMAF kbit/s | LOCMAF bytes | LOCMAF kbit/s | Samples |
|---|
Bytes are the segment as served, without HTTP: for CMAF the styp and every
chunk's moof and mdat header; for LOCMAF the same chunks encoded as LOCMAF
Objects, one MoQ group per segment, without MoQ Object headers. The LOCMAF figures come from the
Eyevinn/locmaf reference encoder, run in this page as
WebAssembly (loading).
This page fetches and walks each segment of the selected track once the player has completed it, and
the same segment number of the other three tracks every fifth segment, to keep the request count low.
kbit/s is the average over the last 10 segments fetched per track.