- Document HF_HOME environment variable for writable cache
- Add systemd service permission guidance for /tmp paths
- Troubleshooting steps for read-only file system errors
JSON stores numbers as strings, so pitch and speed were being returned
as strings from get_user_effects(), causing format string errors like:
'Unknown format code d for object of type str'
Now get_user_effects() explicitly converts:
- pitch to int
- speed to float
This fixes the format string errors when logging or displaying effects.
- Added new audio_effects.py module with pitch shift and speed change
- Pitch range: -12 to +12 semitones (higher = chipmunk, lower = deeper)
- Speed range: 0.5 to 2.0x (higher = faster, lower = slower)
- Maximum 2 active effects per user (performance optimization)
- Added /effects command group:
- /effects list - Shows current effects with descriptions
- /effects set pitch|speed <value> - Apply effects
- /effects reset - Confirmation UI to clear all effects
- Effects persist across restarts in preferences.json
- Updated /voice preview to support optional pitch/speed parameters
- Effects applied in _generate_wav_bytes using librosa
- Added performance warnings when processing takes >1 second
- Updated README with effects documentation
Refactor the TTS handling to support multiple, user-selectable voices. This replaces the previous single-voice system.
Key changes:
- Introduce VoiceManager to handle loading and managing voices from a dedicated oices/ directory.
- Add slash commands (/voice list, /set, /current, /refresh) for users to manage their personal TTS voice.
- Implement on-demand voice loading to improve startup time and memory usage.
- Remove the old ts_handler.py and single voice .wav files in favor of the new system.
- Update configuration to specify a voices directory instead of a single file path.
