MPE and Video

needs verification

The Media Processing Engine (MPE) transforms recorded chapter takes into finished media. A processing run records one attempt, with a saved request, result and success or failure state. Core sends a snapshot so a queued request does not silently change with later edits.

A successful result identifies renditions, subtitles and a poster. A rendition is a playable media version; subtitles carry timed text; the poster is the still image before playback. Core validates the result and registers existing R2 objects, then projects a Video. Registration itself does not copy the output.

The Video stores the playable output, including chapters, duration, layout and file associations. A new projected output supersedes the previous current Video while retaining its record. The LT Work response points to the current ready Video. A ready Video can exist before publication and remain available while a later processing attempt runs.

The development mock queues completion after 30 seconds. Queue delay is not a guarantee of completion at exactly 30 seconds.

Delivery boundary

The player configuration API requires authentication and work visibility. It is not an anonymous website endpoint. Its share URL is currently the configured homepage, not a video's published page. MPE has no checked-out repository or documented service URL in this workspace.

For maintainers: source references

These references require repository access. Everything needed to understand this page is explained above.

• integration contract
• result handling
• Video model
• LT Work resource
• mock processing client
• Video API routes
• integration notes