| commit | author | age | ||
| 564abd | 1 | import enum |
| JK | 2 | import io |
| 3 | ||
| 4 | import fastapi | |
| 5 | import pydantic | |
| 6 | import TTS.api | |
| 7 | ||
| 8 | import config | |
| 9 | ||
| 10 | models = {} | |
| 11 | ||
| 12 | for id, model in config.models.items(): | |
| 13 | models[id] = TTS.api.TTS(model).to("cpu") | |
| 14 | ||
| 15 | ||
| 16 | LanguageModel = enum.Enum('LanguageModel', {k: k for k in models.keys()}) | |
| 17 | ||
| 18 | ||
| 19 | class SynthesizeRequest(pydantic.BaseModel): | |
| 20 | language: LanguageModel | |
| 21 | text: str | |
| 22 | ||
| 23 | ||
| 24 | class SynthesizeResponse(fastapi.Response): | |
| 25 | media_type = 'audio/wav' | |
| 26 | ||
| 27 | ||
| 28 | app = fastapi.FastAPI() | |
| 29 | ||
| 30 | @app.post('/synthesize', response_class=SynthesizeResponse) | |
| 31 | async def synthesize(request: SynthesizeRequest) -> SynthesizeResponse: | |
| 32 | with io.BytesIO() as fp: | |
| 33 | models[request.language.value].tts_to_file(request.text, file_path=fp) | |
| 34 | return SynthesizeResponse(content = fp.getvalue()) | |
