Codex
87 lines
2.6 KiB
Python
87 lines
2.6 KiB
Python
"""
|
||
语音识别 API
|
||
"""
|
||
from fastapi import APIRouter, Depends, File, HTTPException, Query, UploadFile
|
||
from fastapi.responses import Response
|
||
from sqlalchemy.ext.asyncio import AsyncSession
|
||
|
||
from app.database import get_db
|
||
|
||
router = APIRouter(prefix="/speech", tags=["AI-语音"])
|
||
|
||
|
||
@router.post("/transcribe", summary="语音转文字")
|
||
async def transcribe(
|
||
audio: UploadFile = File(..., description="音频文件"),
|
||
language: str = Query("zh", description="语言"),
|
||
provider: str = Query("dashscope", description="提供商"),
|
||
db: AsyncSession = Depends(get_db),
|
||
):
|
||
"""
|
||
语音转文字（ASR）
|
||
|
||
将音频文件转换为文字，支持：
|
||
- dashscope: 阿里云百炼（默认，推荐）
|
||
- openai: OpenAI Whisper
|
||
|
||
支持的音频格式：wav, mp3, webm, pcm, opus
|
||
"""
|
||
from ai_platform.services.speech_service import SpeechService
|
||
|
||
service = SpeechService(db=db)
|
||
await service._resolve_dashscope_api_key()
|
||
audio_content = await audio.read()
|
||
result = service.transcribe(
|
||
audio_file=audio_content,
|
||
language=language,
|
||
provider=provider,
|
||
)
|
||
|
||
if result["success"]:
|
||
return {
|
||
"text": result["text"],
|
||
"duration": result.get("duration", 0),
|
||
}
|
||
else:
|
||
raise HTTPException(status_code=400, detail=result["error"])
|
||
|
||
|
||
@router.post("/tts", summary="文字转语音")
|
||
async def text_to_speech(
|
||
text: str = Query(..., description="要转换的文字"),
|
||
voice: str = Query("sambert-zhichu-v1", description="声音"),
|
||
provider: str = Query("dashscope", description="提供商"),
|
||
db: AsyncSession = Depends(get_db),
|
||
):
|
||
"""
|
||
文字转语音（TTS）
|
||
|
||
将文字转换为语音，返回音频文件
|
||
|
||
DashScope 可用声音：
|
||
- sambert-zhichu-v1: 知厨（男声）
|
||
- sambert-zhimiao-emo-v1: 知妙（女声，带情感）
|
||
- sambert-zhiying-v1: 知莺（女声）
|
||
|
||
OpenAI 可用声音：
|
||
- alloy, echo, fable, onyx, nova, shimmer
|
||
"""
|
||
from ai_platform.services.speech_service import SpeechService
|
||
|
||
service = SpeechService(db=db)
|
||
await service._resolve_dashscope_api_key()
|
||
result = service.text_to_speech(
|
||
text=text,
|
||
voice=voice,
|
||
provider=provider,
|
||
)
|
||
|
||
if result["success"]:
|
||
return Response(
|
||
content=result["audio_data"],
|
||
media_type=result["content_type"],
|
||
headers={"Content-Disposition": 'attachment; filename="speech.wav"'},
|
||
)
|
||
else:
|
||
raise HTTPException(status_code=400, detail=result["error"])
|