Eyevinn / mp4ff / 36253969765

83%

Build Type

push

github

Committed by tobbee

Commit Message

feat(mp4): add experimental paint-model subtitle boxes

Low-latency subtitles are chunked at the video chunk cadence, which today means restating the full document in every chunk even when nothing on screen has changed. These boxes let a chunk say "no change" in eight bytes instead.

The stpc and wvtc sample entries have the same syntax as stpp and wvtt and reuse StppBox and WvttBox with a name field, as the visual sample entries already do. A new 4CC keeps legacy clients from selecting the Representation through the RFC 6381 codecs parameter, since a sample under it is no longer always a document.

ttmn and vttn are 8-byte empty boxes that are a whole sample and mean that the active document or cues continue unchanged, following the emeb precedent of ISO/IEC 23001-18 rather than a zero-size sample, which 14496-30 Section 4.2 forbids. ttmb is a whole sample carrying only the body element of a TTML document, with the head spliced from the first sample of the same fragment; it is therefore not a sync sample.

A sample is one of these boxes only if its first eight bytes are a box header of that type with a size equal to the sample size. A TTML document can never match, since it cannot start with a zero byte. mp4ff-subslister applies that rule and renders the boxes instead of dumping opaque bytes.

The 4CCs are placeholders taken from the design and are not registered with MP4RA.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>

Coverage Stats

119 of 135 new or added lines in 7 files covered. (88.15%)

2 existing lines in 1 file now uncovered.

22350 of 27174 relevant lines covered (82.25%)

0.82 hits per line