Every book, retold for your ears. ReTold turns a plain-text novel into a listening-optimized audiobook – distinct character voices, ambient scene-setting, "till now…" recaps, a world-building intro, and (conservatively) real sound effects in place of throwaway sound-describing sentences.
🔊 Listen to a sample of rendered engine output
| Level | Diagram |
|---|---|
| L1 – core loop (Book → IR → audio) | ![]() |
| L2 – retold engine internals | ![]() |
| L3 – spoiler-safe assistant | ![]() |
Editable .svg sources: demo/architecture/.
retold_engine/– the per-chapter audiobook generator. Everything centers on the IR (retold/schema.py):Book → Chapter → Scene → Segment.talking_system/– the spoiler-safe "narrator" assistant.server/– a thin FastAPI layer that turns the two into a runnable app with a browser UI.
All showcase assets (video, architecture images, sample audio) live in
media/ – see media/README.md.
pip install numpy pydantic pyyaml # ffmpeg must be on PATH for MP3/M4B
python -m retold.pipeline samples/mad_tea_party.txt --offlineSee retold_engine/README.md for full engine docs and
server/README.md for running the app server.


