lenML/ChatTTS-Forge

1import logging2 3from fastapi import HTTPException, Query, Request4from fastapi.responses import StreamingResponse5from pydantic import BaseModel6 7from modules.api.Api import APIManager8from modules.api.impl.handler.TTSHandler import TTSHandler9from modules.api.impl.model.audio_model import AdjustConfig, AudioFormat10from modules.api.impl.model.chattts_model import ChatTTSConfig, InferConfig11from modules.api.impl.model.enhancer_model import EnhancerConfig12from modules.speaker import speaker_mgr13 14logger = logging.getLogger(__name__)15 16 17class XTTS_V2_Settings:18    def __init__(self):19        self.stream_chunk_size = 10020        self.temperature = 0.321        self.speed = 122 23        # TODO: 这两个参数现在用不着...但是其实gpt是可以用的可以考虑增加24        self.length_penalty = 0.525        self.repetition_penalty = 1.026 27        self.top_p = 0.728        self.top_k = 2029        self.enable_text_splitting = True30 31        # 下面是额外配置 xtts_v2 中不包含的，但是本系统需要的32        self.batch_size = 433        self.eos = "[uv_break]"34        self.infer_seed = 4235        self.use_decoder = True36        self.prompt1 = ""37        self.prompt2 = ""38        self.prefix = ""39        self.spliter_threshold = 10040        self.style = ""41 42 43class TTSSettingsRequest(BaseModel):44    # 这个 stream_chunk 现在当作 spliter_threshold 用45    stream_chunk_size: int46    temperature: float47    speed: float48    length_penalty: float49    repetition_penalty: float50    top_p: float51    top_k: int52    enable_text_splitting: bool53 54    batch_size: int = None55    eos: str = None56    infer_seed: int = None57    use_decoder: bool = None58    prompt1: str = None59    prompt2: str = None60    prefix: str = None61    spliter_threshold: int = None62    style: str = None63 64 65class SynthesisRequest(BaseModel):66    text: str67    speaker_wav: str68    language: str69 70 71def setup(app: APIManager):72    XTTSV2 = XTTS_V2_Settings()73 74    @app.get("/v1/xtts_v2/speakers")75    async def speakers():76        spks = speaker_mgr.list_speakers()77        return [78            {79                "name": spk.name,80                "voice_id": spk.id,81                # TODO: 也许可以放一个 "/v1/tts" 接口地址在这里82                "preview_url": "",83            }84            for spk in spks85        ]86 87    @app.post("/v1/xtts_v2/tts_to_audio", response_class=StreamingResponse)88    async def tts_to_audio(request: SynthesisRequest):89        text = request.text90        # speaker_wav 就是 speaker id 。。。91        voice_id = request.speaker_wav92        language = request.language93 94        spk = speaker_mgr.get_speaker_by_id(voice_id) or speaker_mgr.get_speaker(95            voice_id96        )97        if spk is None:98            raise HTTPException(status_code=400, detail="Invalid speaker id")99 100        tts_config = ChatTTSConfig(101            style=XTTSV2.style,102            temperature=XTTSV2.temperature,103            top_k=XTTSV2.top_k,104            top_p=XTTSV2.top_p,105            prefix=XTTSV2.prefix,106            prompt1=XTTSV2.prompt1,107            prompt2=XTTSV2.prompt2,108        )109        infer_config = InferConfig(110            batch_size=XTTSV2.batch_size,111            spliter_threshold=XTTSV2.spliter_threshold,112            eos=XTTSV2.eos,113            seed=XTTSV2.infer_seed,114        )115        adjust_config = AdjustConfig(116            speed_rate=XTTSV2.speed,117        )118        # TODO: support enhancer119        enhancer_config = EnhancerConfig(120            # enabled=params.enhance or params.denoise or False,121            # lambd=0.9 if params.denoise else 0.1,122        )123 124        handler = TTSHandler(125            text_content=text,126            spk=spk,127            tts_config=tts_config,128            infer_config=infer_config,129            adjust_config=adjust_config,130            enhancer_config=enhancer_config,131        )132 133        buffer = handler.enqueue_to_buffer(AudioFormat.mp3)134 135        return StreamingResponse(buffer, media_type="audio/mpeg")136 137    @app.get("/v1/xtts_v2/tts_stream")138    async def tts_stream(139        request: Request,140        text: str = Query(),141        speaker_wav: str = Query(),142        language: str = Query(),143    ):144        # speaker_wav 就是 speaker id 。。。145        voice_id = speaker_wav146 147        spk = speaker_mgr.get_speaker_by_id(voice_id) or speaker_mgr.get_speaker(148            voice_id149        )150        if spk is None:151            raise HTTPException(status_code=400, detail="Invalid speaker id")152 153        tts_config = ChatTTSConfig(154            style=XTTSV2.style,155            temperature=XTTSV2.temperature,156            top_k=XTTSV2.top_k,157            top_p=XTTSV2.top_p,158            prefix=XTTSV2.prefix,159            prompt1=XTTSV2.prompt1,160            prompt2=XTTSV2.prompt2,161        )162        infer_config = InferConfig(163            batch_size=XTTSV2.batch_size,164            spliter_threshold=XTTSV2.spliter_threshold,165            eos=XTTSV2.eos,166            seed=XTTSV2.infer_seed,167        )168        adjust_config = AdjustConfig(169            speed_rate=XTTSV2.speed,170        )171        # TODO: support enhancer172        enhancer_config = EnhancerConfig(173            # enabled=params.enhance or params.denoise or False,174            # lambd=0.9 if params.denoise else 0.1,175        )176 177        handler = TTSHandler(178            text_content=text,179            spk=spk,180            tts_config=tts_config,181            infer_config=infer_config,182            adjust_config=adjust_config,183            enhancer_config=enhancer_config,184        )185 186        async def generator():187            for chunk in handler.enqueue_to_stream(AudioFormat.mp3):188                disconnected = await request.is_disconnected()189                if disconnected:190                    break191 192                yield chunk193 194        return StreamingResponse(generator(), media_type="audio/mpeg")195 196    @app.post("/v1/xtts_v2/set_tts_settings")197    async def set_tts_settings(request: TTSSettingsRequest):198        try:199            if request.stream_chunk_size < 50:200                raise HTTPException(201                    status_code=400, detail="stream_chunk_size must be greater than 0"202                )203            if request.temperature < 0:204                raise HTTPException(205                    status_code=400, detail="temperature must be greater than 0"206                )207            if request.speed < 0:208                raise HTTPException(209                    status_code=400, detail="speed must be greater than 0"210                )211            if request.length_penalty < 0:212                raise HTTPException(213                    status_code=400, detail="length_penalty must be greater than 0"214                )215            if request.repetition_penalty < 0:216                raise HTTPException(217                    status_code=400, detail="repetition_penalty must be greater than 0"218                )219            if request.top_p < 0:220                raise HTTPException(221                    status_code=400, detail="top_p must be greater than 0"222                )223            if request.top_k < 0:224                raise HTTPException(225                    status_code=400, detail="top_k must be greater than 0"226                )227 228            XTTSV2.stream_chunk_size = request.stream_chunk_size229            XTTSV2.spliter_threshold = request.stream_chunk_size230 231            XTTSV2.temperature = request.temperature232            XTTSV2.speed = request.speed233            XTTSV2.length_penalty = request.length_penalty234            XTTSV2.repetition_penalty = request.repetition_penalty235            XTTSV2.top_p = request.top_p236            XTTSV2.top_k = request.top_k237            XTTSV2.enable_text_splitting = request.enable_text_splitting238 239            # TODO: checker240            if request.batch_size:241                XTTSV2.batch_size = request.batch_size242            if request.eos:243                XTTSV2.eos = request.eos244            if request.infer_seed:245                XTTSV2.infer_seed = request.infer_seed246            if request.use_decoder:247                XTTSV2.use_decoder = request.use_decoder248            if request.prompt1:249                XTTSV2.prompt1 = request.prompt1250            if request.prompt2:251                XTTSV2.prompt2 = request.prompt2252            if request.prefix:253                XTTSV2.prefix = request.prefix254            if request.spliter_threshold:255                XTTSV2.spliter_threshold = request.spliter_threshold256            if request.style:257                XTTSV2.style = request.style258 259            return {"message": "Settings successfully applied"}260        except Exception as e:261            if isinstance(e, HTTPException):262                raise e263            logger.error(e)264            raise HTTPException(status_code=500, detail=str(e))265