- Created PiperTTSService with pitch shifting (-12 to +12 semitones)
- Created UnifiedTTSService to manage multiple TTS engines
- Added engine selection endpoints (GET/POST /api/tts/engines)
- Added voices listing endpoint (GET /api/tts/voices)
- Added pitch_adjustments parameter to TTS generation
- Created install_piper.sh for easy Piper installation
- Updated backend README with multi-engine documentation
- Created comprehensive PIPER_TTS_GUIDE.md
- Added TTS_ENGINE environment variable for engine selection
- Updated requirements.txt with piper-tts and pydub
This allows users to choose between Kokoro (high quality) and Piper
(fast with pitch control) TTS engines at runtime.
Updated the speaker detection prompt to explicitly instruct the AI to:
- Identify narrative text and assign it to 'Narrator'
- Detect character dialogue and list character names
- Include BOTH Narrator and characters when story has mixed content
This ensures stories with narrative descriptions get proper voice
assignments for both the narrator and speaking characters.
Example: Story with 'Once upon a time...' narrative and dialogue from
Lily, Benny, Sammy now correctly detects all 4 roles.
- Added ScriptRequest, ScriptResponse, and ScriptSegment Pydantic models
- Created /api/tts/script POST endpoint in main.py
- Integrated with TTSService to generate script segments using AI
- Added generate_script_segments method to TTSService
- Returns segments with unique IDs for frontend tracking
- Handles errors gracefully when AI segmentation fails
- Falls back to pattern-based segmentation when Ollama unavailable
- Added comprehensive tests for the endpoint
- Validates: Requirements 15.1, 15.2
- Added segment_text() method to OllamaService for AI-based text splitting
- OllamaService now annotates each text segment with the appropriate speaker
- Updated TTSService to use AI segmentation when available
- Falls back to pattern-based segmentation when Ollama unavailable
- AI intelligently assigns dialogue to speakers and narrative to Narrator
- Added comprehensive test suite for AI segmentation
- Verified AI correctly splits unmarked dialogue between multiple speakers
This makes the tool much more powerful - users can now input plain narrative
text without explicit role markers and the AI will intelligently assign voices
to different characters and the narrator.
- Changed model from 'llama3:8b' to 'llama3:latest' to match available models
- Verified Ollama integration is working correctly
- AI detection now successfully identifies speakers in unmarked text
- Sample up to 4 chunks of 5000 characters from different text positions
- Distribute samples evenly across beginning, middle, and end of text
- Deduplicate speakers across all samples for comprehensive detection
- Log individual sample results for debugging
- Handle short texts efficiently (single sample)
- Updated tests to verify multi-sample behavior
This ensures speakers introduced anywhere in long texts are detected,
not just those appearing in the first 2000 characters.
- Add POST /api/tts endpoint with full request validation
- Validate text, language code, speed range, and roles list
- Integrate TTSService to generate multi-voice audiobooks
- Fix Kokoro pipeline generator handling to properly collect audio
- Stream audio response with correct Content-Type (audio/wav)
- Add Content-Disposition header with filename (audiobook.wav)
- Handle errors gracefully with appropriate HTTP status codes
- Add comprehensive test suite for endpoint validation
Requirements: 6.1, 6.7, 6.8, 9.5
- Created TTSService class with pipeline caching per language
- Implemented text segmentation to identify role markers (colon and bracket patterns)
- Implemented segment synthesis with voice assignment
- Implemented audio concatenation at 24 kHz sample rate
- Added fallback to default voice (af_bella) for unmapped roles
- Supports language codes: a, b, f, j
- Handles long text by chunking on sentence boundaries
- Added comprehensive unit tests (13 tests, all passing)
Requirements: 10.1, 10.3, 10.4, 6.2, 6.3, 6.4, 6.5, 6.6
- Created RoleDetector class with colon and bracket pattern extraction
- Implemented role deduplication logic
- Added fallback to default 'Narrator' role when no patterns found
- Created /api/tts/roles endpoint using RoleDetector
- Added comprehensive test suite for role detection
- All tests passing (7/7)
Requirements: 2.1, 2.2, 2.3, 2.4, 2.5, 2.6
- Created backend directory with FastAPI project structure
- Created frontend directory with React + TypeScript + Vite
- Configured Python virtual environment and installed dependencies (fastapi, uvicorn, kokoro, soundfile, python-multipart)
- Verified espeak-ng system dependency is installed
- Configured Node.js project with TypeScript, React, and Vite
- Set up Tailwind CSS for styling
- Set up environment variable configuration for both projects
- Created comprehensive README documentation
Requirements: 9.6, 10.2, 11.1
