GPT-SoVITS/api_v3.py

"""
# WebAPI文档

` python api_v2.py -a 127.0.0.1 -p 9880 -c GPT_SoVITS/configs/tts_infer.yaml `

## 执行参数:
    `-a` - `绑定地址, 默认"127.0.0.1"`
    `-p` - `绑定端口, 默认9880`
    `-c` - `TTS配置文件路径, 默认"GPT_SoVITS/configs/tts_infer.yaml"`

## 调用:

### 推理

endpoint: `/tts`
GET:
```
http://127.0.0.1:9880/tts?text=先帝创业未半而中道崩殂，今天下三分，益州疲弊，此诚危急存亡之秋也。&text_lang=zh&ref_audio_path=archive_jingyuan_1.wav&prompt_lang=zh&prompt_text=我是「罗浮」云骑将军景元。不必拘谨，「将军」只是一时的身份，你称呼我景元便可&text_split_method=cut5&batch_size=1&media_type=wav&streaming_mode=true
```

POST:
```json
{
    "text": "",                   # str.(required) text to be synthesized
    "text_lang: "",               # str.(required) language of the text to be synthesized
    "ref_audio_path": "",         # str.(required) reference audio path
    "aux_ref_audio_paths": [],    # list.(optional) auxiliary reference audio paths for multi-speaker tone fusion
    "prompt_text": "",            # str.(optional) prompt text for the reference audio
    "prompt_lang": "",            # str.(required) language of the prompt text for the reference audio
    "top_k": 15,                  # int. top k sampling
    "top_p": 1,                   # float. top p sampling
    "temperature": 1,             # float. temperature for sampling
    "text_split_method": "cut5",  # str. text split method, see text_segmentation_method.py for details.
    "batch_size": 1,              # int. batch size for inference
    "batch_threshold": 0.75,      # float. threshold for batch splitting.
    "split_bucket": True,         # bool. whether to split the batch into multiple buckets.
    "speed_factor":1.0,           # float. control the speed of the synthesized audio.
    "fragment_interval":0.3,      # float. to control the interval of the audio fragment.
    "seed": -1,                   # int. random seed for reproducibility.
    "parallel_infer": True,       # bool. whether to use parallel inference.
    "repetition_penalty": 1.35,   # float. repetition penalty for T2S model.
    "sample_steps": 32,           # int. 仅 v3/v4 vocoder 路径使用；当前 v2/v2ProPlus 主线可忽略。
    "super_sampling": False,      # bool. 仅 v3/v4 路径使用；不属于当前 v2/v2ProPlus 正式支持目标。
    "streaming_mode": False,      # bool or int. return audio chunk by chunk.T he available options are: 0,1,2,3 or True/False (0/False: Disabled | 1/True: Best Quality, Slowest response speed (old version streaming_mode) | 2: Medium Quality, Slow response speed | 3: Lower Quality, Faster response speed )
    "overlap_length": 2,          # int. overlap length of semantic tokens for streaming mode.
    "min_chunk_length": 16,       # int. The minimum chunk length of semantic tokens for streaming mode. (affects audio chunk size)
}
```

RESP:
成功: 直接返回 wav 音频流， http code 200
失败: 返回包含错误信息的 json, http code 400

### 命令控制

endpoint: `/control`

command:
"restart": 重新运行
"exit": 结束运行

GET:
```
http://127.0.0.1:9880/control?command=restart
```
POST:
```json
{
    "command": "restart"
}
```

RESP: 无


### 切换GPT模型

endpoint: `/set_gpt_weights`

GET:
```
http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_SoVITS/pretrained_models/s1v3.ckpt
```
RESP:
成功: 返回"success", http code 200
失败: 返回包含错误信息的 json, http code 400


### 切换Sovits模型

endpoint: `/set_sovits_weights`

GET:
```
http://127.0.0.1:9880/set_sovits_weights?weights_path=GPT_SoVITS/pretrained_models/v2Pro/s2Gv2ProPlus.pth
```

RESP:
成功: 返回"success", http code 200
失败: 返回包含错误信息的 json, http code 400

"""

import os
import sys
import traceback
from typing import List, Union

now_dir = os.getcwd()
sys.path.append(now_dir)
sys.path.append("%s/GPT_SoVITS" % (now_dir))

from runtime_preload import preload_text_runtime_deps

preload_text_runtime_deps()

import argparse
import signal
from fastapi import FastAPI, Response
from fastapi.responses import StreamingResponse, JSONResponse
import uvicorn
from tools.i18n.i18n import I18nAuto
from GPT_SoVITS.TTS_infer_pack.TTS import TTS, TTS_Config
from GPT_SoVITS.TTS_infer_pack.unified_engine import RuntimeControlCallbacks, UnifiedTTSEngine
from GPT_SoVITS.TTS_infer_pack.text_segmentation_method import get_method_names as get_cut_method_names
from pydantic import BaseModel

# print(sys.path)
i18n = I18nAuto()
cut_method_names = get_cut_method_names()

parser = argparse.ArgumentParser(description="GPT-SoVITS api")
parser.add_argument("-c", "--tts_config", type=str, default="GPT_SoVITS/configs/tts_infer.yaml", help="tts_infer路径")
parser.add_argument("-a", "--bind_addr", type=str, default="127.0.0.1", help="default: 127.0.0.1")
parser.add_argument("-p", "--port", type=int, default="9880", help="default: 9880")
args = parser.parse_args()
config_path = args.tts_config
# device = args.device
port = args.port
host = args.bind_addr
argv = sys.argv

if config_path in [None, ""]:
    config_path = "GPT-SoVITS/configs/tts_infer.yaml"

tts_config = TTS_Config(config_path)
print(tts_config)
tts_pipeline = TTS(tts_config)
tts_engine = UnifiedTTSEngine(
    tts_pipeline,
    cut_method_names=cut_method_names,
    control_callbacks=RuntimeControlCallbacks(
        restart=lambda: os.execl(sys.executable, sys.executable, *argv),
        exit=lambda: os.kill(os.getpid(), signal.SIGTERM),
    ),
)

APP = FastAPI()


class TTS_Request(BaseModel):
    text: str = None
    text_lang: str = None
    ref_audio_path: str = None
    aux_ref_audio_paths: list = None
    prompt_lang: str = None
    prompt_text: str = ""
    top_k: int = 15
    top_p: float = 1
    temperature: float = 1
    text_split_method: str = "cut5"
    batch_size: int = 1
    batch_threshold: float = 0.75
    split_bucket: bool = True
    speed_factor: float = 1.0
    fragment_interval: float = 0.3
    seed: int = -1
    media_type: str = "wav"
    streaming_mode: Union[bool, int] = False
    parallel_infer: bool = True
    repetition_penalty: float = 1.35
    sample_steps: int = 32
    super_sampling: bool = False
    overlap_length: int = 2
    min_chunk_length: int = 16


class Scheduler_Debug_Request_Item(BaseModel):
    request_id: str | None = None
    text: str
    text_lang: str
    ref_audio_path: str
    prompt_lang: str
    prompt_text: str = ""
    top_k: int = 15
    top_p: float = 1
    temperature: float = 1
    repetition_penalty: float = 1.35
    early_stop_num: int = -1
    ready_step: int = 0


class Scheduler_Debug_Request(BaseModel):
    requests: List[Scheduler_Debug_Request_Item]
    max_steps: int = 1500
    seed: int = -1


class Scheduler_Submit_Request(BaseModel):
    request_id: str | None = None
    text: str
    text_lang: str
    ref_audio_path: str
    prompt_lang: str
    prompt_text: str = ""
    top_k: int = 15
    top_p: float = 1
    temperature: float = 1
    repetition_penalty: float = 1.35
    early_stop_num: int = -1
    speed_factor: float = 1.0
    sample_steps: int = 32
    media_type: str = "wav"
    timeout_sec: float = 30.0


def _lower_or_none(value: str | None) -> str | None:
    return value.lower() if isinstance(value, str) else value


async def tts_scheduler_debug_handle(request: Scheduler_Debug_Request):
    try:
        result = await tts_engine.run_scheduler_debug(
            request_items=[item.dict() for item in request.requests],
            max_steps=int(request.max_steps),
            seed=int(request.seed),
        )
        return JSONResponse(status_code=200, content=result.payload)
    except Exception as e:
        return JSONResponse(
            status_code=400,
            content={"message": "scheduler debug failed", "Exception": str(e)},
        )


async def tts_scheduler_submit_handle(request: Scheduler_Submit_Request):
    try:
        result = await tts_engine.run_scheduler_submit(request.dict())
        return Response(result.audio_bytes, media_type=result.media_type, headers=result.headers)
    except Exception as e:
        return JSONResponse(
            status_code=400,
            content={"message": "scheduler submit failed", "Exception": str(e)},
        )


async def tts_handle(req: dict):
    """
    Text to speech handler.

    Args:
        req (dict):
            {
                "text": "",                   # str.(required) text to be synthesized
                "text_lang: "",               # str.(required) language of the text to be synthesized
                "ref_audio_path": "",         # str.(required) reference audio path
                "aux_ref_audio_paths": [],    # list.(optional) auxiliary reference audio paths for multi-speaker tone fusion
                "prompt_text": "",            # str.(optional) prompt text for the reference audio
                "prompt_lang": "",            # str.(required) language of the prompt text for the reference audio
                "top_k": 15,                  # int. top k sampling
                "top_p": 1,                   # float. top p sampling
                "temperature": 1,             # float. temperature for sampling
                "text_split_method": "cut5",  # str. text split method, see text_segmentation_method.py for details.
                "batch_size": 1,              # int. batch size for inference
                "batch_threshold": 0.75,      # float. threshold for batch splitting.
                "split_bucket": True,         # bool. whether to split the batch into multiple buckets.
                "speed_factor":1.0,           # float. control the speed of the synthesized audio.
                "fragment_interval":0.3,      # float. to control the interval of the audio fragment.
                "seed": -1,                   # int. random seed for reproducibility.
                "parallel_infer": True,       # bool. whether to use parallel inference.
                "repetition_penalty": 1.35,   # float. repetition penalty for T2S model.
                "sample_steps": 32,           # int. number of sampling steps for VITS model V3.
                "super_sampling": False,      # bool. only for v3/v4; not part of current v2/v2ProPlus mainline.
                "streaming_mode": False,      # bool or int. return audio chunk by chunk.T he available options are: 0,1,2,3 or True/False (0/False: Disabled | 1/True: Best Quality, Slowest response speed (old version streaming_mode) | 2: Medium Quality, Slow response speed | 3: Lower Quality, Faster response speed )
                "overlap_length": 2,          # int. overlap length of semantic tokens for streaming mode.
                "min_chunk_length": 16,       # int. The minimum chunk length of semantic tokens for streaming mode. (affects audio chunk size)
            }
    returns:
        StreamingResponse: audio stream response.
    """

    try:
        result = await tts_engine.run_direct_tts_async(req)
        if result.streaming:
            return StreamingResponse(result.audio_generator, media_type=f"audio/{result.media_type}")
        return Response(result.audio_bytes, media_type=f"audio/{result.media_type}")
    except Exception as e:
        return JSONResponse(status_code=400, content={"message": "tts failed", "Exception": str(e)})


@APP.get("/control")
async def control(command: str = None):
    if command is None:
        return JSONResponse(status_code=400, content={"message": "command is required"})
    try:
        tts_engine.handle_control(command)
        return JSONResponse(status_code=200, content={"message": "success"})
    except Exception as e:
        return JSONResponse(status_code=400, content={"message": "control failed", "Exception": str(e)})


@APP.get("/tts")
async def tts_get_endpoint(
    text: str = None,
    text_lang: str = None,
    ref_audio_path: str = None,
    aux_ref_audio_paths: list = None,
    prompt_lang: str = None,
    prompt_text: str = "",
    top_k: int = 15,
    top_p: float = 1,
    temperature: float = 1,
    text_split_method: str = "cut5",
    batch_size: int = 1,
    batch_threshold: float = 0.75,
    split_bucket: bool = True,
    speed_factor: float = 1.0,
    fragment_interval: float = 0.3,
    seed: int = -1,
    media_type: str = "wav",
    parallel_infer: bool = True,
    repetition_penalty: float = 1.35,
    sample_steps: int = 32,
    super_sampling: bool = False,
    streaming_mode: Union[bool, int] = False,
    overlap_length: int = 2,
    min_chunk_length: int = 16,
):
    req = {
        "text": text,
        "text_lang": _lower_or_none(text_lang),
        "ref_audio_path": ref_audio_path,
        "aux_ref_audio_paths": aux_ref_audio_paths,
        "prompt_text": prompt_text,
        "prompt_lang": _lower_or_none(prompt_lang),
        "top_k": top_k,
        "top_p": top_p,
        "temperature": temperature,
        "text_split_method": text_split_method,
        "batch_size": int(batch_size),
        "batch_threshold": float(batch_threshold),
        "speed_factor": float(speed_factor),
        "split_bucket": split_bucket,
        "fragment_interval": fragment_interval,
        "seed": seed,
        "media_type": media_type,
        "streaming_mode": streaming_mode,
        "parallel_infer": parallel_infer,
        "repetition_penalty": float(repetition_penalty),
        "sample_steps": int(sample_steps),
        "super_sampling": super_sampling,
        "overlap_length": int(overlap_length),
        "min_chunk_length": int(min_chunk_length),
    }
    return await tts_handle(req)


@APP.post("/tts")
async def tts_post_endpoint(request: TTS_Request):
    req = request.dict()
    return await tts_handle(req)


@APP.post("/tts_scheduler_debug")
async def tts_scheduler_debug_endpoint(request: Scheduler_Debug_Request):
    return await tts_scheduler_debug_handle(request)


@APP.post("/tts_scheduler_submit")
async def tts_scheduler_submit_endpoint(request: Scheduler_Submit_Request):
    return await tts_scheduler_submit_handle(request)


@APP.get("/tts_scheduler_state")
async def tts_scheduler_state_endpoint():
    return JSONResponse(status_code=200, content=tts_engine.get_runtime_state())


@APP.get("/set_refer_audio")
async def set_refer_aduio(refer_audio_path: str = None):
    try:
        payload = tts_engine.set_refer_audio(refer_audio_path)
    except Exception as e:
        return JSONResponse(status_code=400, content={"message": "set refer audio failed", "Exception": str(e)})
    return JSONResponse(status_code=200, content=payload)


# @APP.post("/set_refer_audio")
# async def set_refer_aduio_post(audio_file: UploadFile = File(...)):
#     try:
#         # 检查文件类型，确保是音频文件
#         if not audio_file.content_type.startswith("audio/"):
#             return JSONResponse(status_code=400, content={"message": "file type is not supported"})

#         os.makedirs("uploaded_audio", exist_ok=True)
#         save_path = os.path.join("uploaded_audio", audio_file.filename)
#         # 保存音频文件到服务器上的一个目录
#         with open(save_path , "wb") as buffer:
#             buffer.write(await audio_file.read())

#         tts_pipeline.set_ref_audio(save_path)
#     except Exception as e:
#         return JSONResponse(status_code=400, content={"message": f"set refer audio failed", "Exception": str(e)})
#     return JSONResponse(status_code=200, content={"message": "success"})


@APP.get("/set_gpt_weights")
async def set_gpt_weights(weights_path: str = None):
    try:
        payload = tts_engine.set_gpt_weights(weights_path)
    except Exception as e:
        return JSONResponse(status_code=400, content={"message": "change gpt weight failed", "Exception": str(e)})
    return JSONResponse(status_code=200, content=payload)


@APP.get("/set_sovits_weights")
async def set_sovits_weights(weights_path: str = None):
    try:
        payload = tts_engine.set_sovits_weights(weights_path)
    except Exception as e:
        return JSONResponse(status_code=400, content={"message": "change sovits weight failed", "Exception": str(e)})
    return JSONResponse(status_code=200, content=payload)


if __name__ == "__main__":
    try:
        if host == "None":  # 在调用时使用 -a None 参数，可以让api监听双栈
            host = None
        uvicorn.run(app=APP, host=host, port=port, workers=1)
    except Exception:
        traceback.print_exc()
        os.kill(os.getpid(), signal.SIGTERM)
        exit(0)