Project Case Study
WakeTheBook
Local-first audiobook studio for EPUB and PDF files
A local-first full-stack tool that turns EPUB and PDF books into narrated audiobooks with chapter review, voice profiles, and resumable rendering. Runs local TTS models so source files and audio generation stay on-device.
Metrics
Local-first, no cloud audio generation
Rendering model
5-step import → extract → review → render
Workflow
XTTS v2 + VoxCPM2
Voice stack
01
Overview
WakeTheBook was built as a local audiobook production tool, not just a file converter. It gives users one guided flow to import a book, inspect extracted chapters, adjust the structure, choose a voice, and render final audio without sending source files to external services.
02
Key Highlights
- Supports EPUB and PDF import with extraction, chapter detection, and cleanup before audio rendering.
- Lets the user review, edit, split, merge, or skip chapters before committing to the final render.
- Uses local voice workflows with XTTS voice cloning and VoxCPM2 support instead of a cloud TTS dependency.
- Handles rendering as resumable chapter jobs, so failures do not force a full restart.
03
Technical Notes
- FastAPI backend orchestrates ingest, extraction, analysis, voice handling, render jobs, and filesystem-backed project storage.
- React + TypeScript frontend wraps the pipeline in a 5-step wizard and output view for non-technical use.
- SQLite stores project state while generated files live under per-project directories for easy inspection and recovery.
- The app keeps the whole processing loop local, including voice previews, chapter audio, manifests, and logs.
04
Results
- Covers file ingest, chapter cleanup, voice selection, local inference, and long-running render jobs in one product flow.
- Runs without cloud audio generation, keeping source files local and reducing operating cost.
- Shows ownership across product flow, backend orchestration, frontend job UX, and local storage design.