Reusable voice reader active
Reusable draft listening with canonical reference clips and live word tracking
ECHO is a browser-delivered draft-listening surface for hearing writing back through reusable voices seeded by canonical MP3 reference clips.
The current app surface consolidates ECHO, Ariel, and Voice11 under one route: /echo/.
It now runs as a reusable voice reader and reference deck: import file or paste text -> choose a profile -> play, pause, or stop with live highlighting.
The profiles shape pacing and tonal intent while the live readback engine sends the selected canonical MP3 reference clip to the voice worker. When FAL_KEY is configured, that worker uses the clip as the reusable voice base for clone-capable synthesis. The routed MP3 bank now lives in public/echo/ and resolves explicit English and French reference clips for ECHO, Ariel, and Voice11 as the canonical voice source for those routes.
The current work is focused on revision through listening: hearing cadence, awkward phrasing, and sentence-level rhythm without sending text to a backend. This build uses local document extraction plus reference-clip-driven voice generation, with clone synthesis requiring a configured voice backend.
Source provenance for this import pass:
D:\softinfo\ECHO