# feat/ssfm-v30: Add ssfm-v30 Model Support
## Summary
Add ssfm-v30 model support to the MCP server. Includes new emotion presets (whisper, toneup, tonedown), Smart Mode for context-aware emotion inference, and enhanced voice filtering via V2 Voices API.
## Changes
### Features
- **ssfm-v30 model support**: Add SSFM_V30 to TTSModel enum
- **New emotion presets**: whisper, toneup, tonedown (ssfm-v30 only)
- **Preset Mode**: `PresetPrompt` model for explicit emotion control
- **Smart Mode**: `SmartPrompt` model for context-aware emotion inference
- `previous_text`: Previous context text
- `next_text`: Next context text
- **V2 Voices API**: Gender/age filtering support
- `GenderEnum`: male, female
- `AgeEnum`: child, teen, young_adult, middle_aged, senior
- **Default model change**: ssfm-v21 → ssfm-v30
### Bug Fixes
- Fix output file extension to correctly use `audio_format` parameter
### Documentation
- knowledge.py: Add ssfm-v30 model description and features
- knowledge.py: Add Smart Mode usage examples (Python, JavaScript, cURL)
- knowledge.py: Document V2 Voices API specification and filtering
- knowledge.py: Update supported language count (27 → 37)
- README.md: Add supported models table
- README.md: Add ssfm-v30 features section
- README.md: Update feature implementation status
## Changed Files
| File | Changes |
|------|---------|
| `app/server.py` | ssfm-v30 model, new emotion presets, PresetPrompt/SmartPrompt, V2 API support |
| `app/knowledge.py` | ssfm-v30 documentation, Smart Mode examples, V2 API specification |
| `README.md` | Models table, feature status, ssfm-v30 features section |
## Key Code Changes
### New Pydantic Models
```python
class PresetPrompt(BaseModel):
"""Preset-based emotion control for ssfm-v30 model."""
emotion_type: EmotionType = Field(default=EmotionType.PRESET)
emotion_preset: EmotionEnum = Field(default=EmotionEnum.NORMAL)
emotion_intensity: float = Field(default=1.0, ge=0.0, le=2.0)
class SmartPrompt(BaseModel):
"""Context-aware emotion inference for ssfm-v30 model."""
emotion_type: EmotionType = Field(default=EmotionType.SMART)
previous_text: str | None = Field(default=None)
next_text: str | None = Field(default=None)
```
### V2 Voices API Filtering
```python
async def get_voices(
model: str | None = None,
gender: str | None = None, # male, female
age: str | None = None, # child, teen, young_adult, middle_aged, senior
) -> dict:
```
### Extended text_to_speech Parameters
```python
async def text_to_speech(
voice_id: str,
text: str,
model: str = TTSModel.SSFM_V30.value, # Default changed
emotion_type: str = "preset", # New parameter
emotion_preset: str = EmotionEnum.NORMAL.value,
emotion_intensity: float = 1.0,
previous_text: str | None = None, # Smart Mode
next_text: str | None = None, # Smart Mode
...
) -> str:
```
## Testing
- [ ] get_voices: Verify V2 API calls and filtering
- [ ] text_to_speech: Verify Preset Mode emotion generation
- [ ] text_to_speech: Verify Smart Mode context inference
- [ ] Verify new emotion presets (whisper, toneup, tonedown)
## Related Issues
- MCP server update for ssfm-v30 model release
