ENGINEERING INDEX / ENTRY 08
CASE 08 / MEDIA SYSTEMS / SOURCE CONTINUITY
The audio still knows where it came from.
Narration is more useful when a listener can return to the original passage instead of receiving one long, contextless audio file. Reader's forge builds chapter metadata and a passage timeline from source chunks and measured audio durations.
How can a listening edition preserve a route back to the source text and chapter boundaries?
Where the handoffs happen.
The timeline retains chunk, page, heading, and source text references. A measured M4B test uses a synthetic sentence engine and FFmpeg to verify chapter/audio packaging without claiming a full real-book narration run.
- 01Extract source chunks
- 02Narrate passages
- 03Measure audio
- 04Build chapter and text timeline
What the record earns.
15 Reader forge tests passed in the latest focused run.
- Observed receipt
- On September 25, 2026, all 15 focused Reader forge tests passed, including source-linked passage timing, chapter grouping, restart controls, and a measured M4B build using a test audio engine.
- Source and scope
- Twilight Reader forge implementation and focused forge tests in the current working tree. The Reader files inspected here are untracked, so the repository HEAD is not the revision containing this work.
What this does not prove.
A test audio engine and a short fixture prove the forge's local behavior, not a completed full-book render with a production voice, audiobook rights, or mobile playback quality.
Where the work goes next.
Capture a rights-cleared short source-to-listening demonstration with a real voice render, timeline, and player handoff.