The challenge
Turning a long document into an audiobook usually means moving between text cleanup, speech generation, and audio editing tools. I wanted a private workflow that could handle all three on one computer.
// CASE STUDY 04
LOCAL-FIRST AUDIOBOOK STUDIO
BookVoice
A Windows desktop application for converting long-form documents into locally generated narration and shaping the result in a multitrack editor.
- Role
- Desktop application development
- Platform
- Windows desktop application
- Core stack
- Electron · Python · Local TTS
// 01 / THE STORY
Why this was built.
The approach
I split generation by chapter so long jobs can show progress, recover partial work, and be revised. An Electron interface coordinates a local Python speech worker and a three-track editor for narration, ambience, and effects.
What it delivers
BookVoice can take a document through local narration and audio arrangement to an exportable MP3. After the initial model download, the core workflow runs without an account or hosted backend.
// 02
Key capabilities
From raw documents to a structured, mixed, and exportable audiobook.
- Import PDF, TXT, Markdown, or pasted text
- Normalize text and detect editable chapter structures
- Generate narration locally with Piper or Kokoro
- Install the selected Python runtime, engine, and voice model only when needed
- Process chapters sequentially with progress, cancellation, and partial recovery
- Preview and save individual chapters or a combined MP3 audiobook
- Arrange narration, ambience, and effects across three timeline tracks
- Move, trim, split, fade, loop, and adjust clip volume
- Apply narration-aware ducking, resampling, and peak normalization
- Navigate with waveform previews, a draggable playhead, and timeline zoom
- Save reusable .bookvoice editor projects with undo and redo history
- Search Freesound with preview caching, license metadata, and attribution export
// 03
Architecture
A sandboxed Electron interface coordinating long-running local Python audio work.
- Desktop
- Electron, Vite, ES2022 JavaScript, HTML, and CSS
- Processing
- Persistent local Python worker using JSON over stdin and stdout
- Speech
- Piper and Kokoro local text-to-speech engines
- Documents
- PDF.js plus TXT, Markdown, and pasted-text import
- Audio
- NumPy, SoundFile, waveform analysis, offline mixing, and MP3 export
- Distribution
- Electron Builder and a guided Windows NSIS installer
// DESKTOP PRODUCT DEVELOPMENT
Let's work together