zhaojie/tts_trans
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: API文档、文本自动分段、音色配置、批量并发

Browse Source 
- 新增 API.md 完整接口文档
- 智能文本分段：长文本按段落/句子/标点边界自动切分（≤2000字/段），逐段TTS后ffmpeg拼接
- /api/tts 支持 voice 参数指定音色
- httpTts JSON 配置增加 style 和 voice 字段
- 批量生成改用并发（Semaphore 3路）
- 新增 /health 健康检查端点
- TTS 试听前端增加音色输入
- 清理 import，修复端口不一致
This commit is contained in:
sunruiling
2026-03-27 14:37:43 +08:00
parent 82624afe68
commit 30544f7f42
6 changed files with 680 additions and 47 deletions
Download Patch File Download Diff File Expand all files Collapse all files

236
app/main.py
View File

@@ -16,11 +16,12 @@ from pathlib import Path
import httpx
from fastapi import FastAPI, HTTPException, Query, Request
from fastapi.responses import FileResponse, HTMLResponse, JSONResponse
from fastapi.responses import FileResponse, HTMLResponse, JSONResponse, Response
from fastapi.staticfiles import StaticFiles
from sqlalchemy import Column, Integer, String, Text, DateTime, func, select, delete
from sqlalchemy.ext.asyncio import create_async_engine, AsyncSession, async_sessionmaker
from sqlalchemy.orm import DeclarativeBase
from urllib.parse import parse_qs, unquote
import config
@@ -33,6 +34,83 @@ logging.basicConfig(
)
logger = logging.getLogger("tts-service")
# ── Text Segmentation ─────────────────────────────────────────────────────
# MiMo TTS 单次请求文本上限（保守值，实际约 5000）
MAX_CHUNK_CHARS = 2000
# 分割优先级: 段落 > 句子 > 逗号/分号 > 按长度硬切
_SEGMENT_PATTERNS = [
"\n\n", # 段落
"\n", # 换行
"。", "！", "？", "…", # 中文句末
".", "!", "?", # 英文句末
"；", ";", # 分号
"，", ",", # 逗号
]
def split_text(text: str, max_chars: int = MAX_CHUNK_CHARS) -> list[str]:
"""
智能分段：尽量在自然边界（段落/句子/标点）处切分，
保证每段不超过 max_chars 字符。
"""
text = text.strip()
if len(text) <= max_chars:
return [text]
chunks: list[str] = []
remaining = text
while remaining:
if len(remaining) <= max_chars:
chunks.append(remaining)
break
# 在 max_chars 范围内找最佳切割点
window = remaining[:max_chars]
cut_pos = -1
for sep in _SEGMENT_PATTERNS:
idx = window.rfind(sep)
if idx > 0:
cut_pos = idx + len(sep)
break
if cut_pos <= 0:
# 没找到任何分隔符，硬切
cut_pos = max_chars
chunk = remaining[:cut_pos].strip()
if chunk:
chunks.append(chunk)
remaining = remaining[cut_pos:].strip()
return chunks
# ── Audio Concatenation ───────────────────────────────────────────────────
def concat_mp3_files(mp3_paths: list[str], output_path: str):
"""用 ffmpeg 将多个 MP3 文件拼接为一个"""
# 创建 ffmpeg concat 文件列表
list_path = output_path + ".concat_list.txt"
with open(list_path, "w") as f:
for p in mp3_paths:
f.write(f"file '{p}'\n")
try:
result = subprocess.run(
["ffmpeg", "-y", "-f", "concat", "-safe", "0", "-i", list_path,
"-codec:a", "libmp3lame", "-qscale:a", "2", output_path],
capture_output=True, text=True,
)
if result.returncode != 0:
raise RuntimeError(f"ffmpeg 拼接失败: {result.stderr[:300]}")
finally:
os.remove(list_path)
# ── Database ──────────────────────────────────────────────────────────────
engine = create_async_engine(config.DATABASE_URL, echo=False)
@@ -69,16 +147,17 @@ class Chapter(Base):
# ── TTS Service ───────────────────────────────────────────────────────────
async def call_mimo_tts(text: str, style: str = "") -> bytes:
async def call_mimo_tts(text: str, style: str = "", voice: str = "") -> bytes:
"""调用小米 MiMo TTS API，返回 WAV 音频字节"""
if not config.MIMO_API_KEY:
raise HTTPException(500, "MIMO_API_KEY 未配置，请设置环境变量")
content = f"<style>{style}</style>{text}" if style else text
use_voice = voice or config.MIMO_VOICE
payload = {
"model": config.MIMO_TTS_MODEL,
"audio": {"format": "wav", "voice": config.MIMO_VOICE},
"audio": {"format": "wav", "voice": use_voice},
"messages": [{"role": "assistant", "content": content}],
}
@@ -125,7 +204,7 @@ def wav_to_mp3(wav_path: str, mp3_path: str):
async def generate_chapter_audio(chapter_id_str: str):
"""为指定章节生成音频（WAV → MP3）"""
"""为指定章节生成音频（支持长文本自动分段拼接）"""
async with async_session() as db:
result = await db.execute(select(Chapter).where(Chapter.chapter_id == chapter_id_str))
chapter = result.scalar_one_or_none()
@@ -145,20 +224,41 @@ async def generate_chapter_audio(chapter_id_str: str):
audio_dir = Path(config.AUDIO_DIR) / chapter.book_id
audio_dir.mkdir(parents=True, exist_ok=True)
wav_path = str(audio_dir / f"{chapter.chapter_id}.wav")
mp3_path = str(audio_dir / f"{chapter.chapter_id}.mp3")
chunks = split_text(chapter.text_content)
# MiMo TTS 生成 WAV
wav_bytes = await call_mimo_tts(chapter.text_content)
with open(wav_path, "wb") as f:
f.write(wav_bytes)
if len(chunks) == 1:
# 单段：直接生成
wav_bytes = await call_mimo_tts(chapter.text_content)
wav_path = str(audio_dir / f"{chapter.chapter_id}.wav")
with open(wav_path, "wb") as f:
f.write(wav_bytes)
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, mp3_path)
os.remove(wav_path)
else:
# 多段：逐段生成 → 拼接
logger.info(f"章节 {chapter_id_str}: 文本 {len(chapter.text_content)} 字, 分 {len(chunks)} 段生成")
tmp_mp3_paths = []
for i, chunk in enumerate(chunks):
wav_bytes = await call_mimo_tts(chunk)
tmp_id = f"{chapter.chapter_id}_part{i}"
wav_path = str(audio_dir / f"{tmp_id}.wav")
tmp_mp3 = str(audio_dir / f"{tmp_id}.mp3")
# WAV → MP3
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, mp3_path)
with open(wav_path, "wb") as f:
f.write(wav_bytes)
# 删除 WAV 源文件，只保留 MP3
os.remove(wav_path)
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, tmp_mp3)
os.remove(wav_path)
tmp_mp3_paths.append(tmp_mp3)
# 拼接
await loop.run_in_executor(None, concat_mp3_files, tmp_mp3_paths, mp3_path)
for p in tmp_mp3_paths:
os.remove(p)
logger.info(f"章节 {chapter_id_str}: {len(chunks)} 段拼接完成")
chapter.audio_file = mp3_path
chapter.status = "ready"
@@ -166,6 +266,7 @@ async def generate_chapter_audio(chapter_id_str: str):
except Exception as e:
chapter.status = "error"
chapter.error_msg = str(e)[:500]
logger.error(f"章节 {chapter_id_str} 生成失败: {e}")
await db.commit()
@@ -184,6 +285,12 @@ async def lifespan(app: FastAPI):
app = FastAPI(title="TTS Book Service", lifespan=lifespan)
@app.get("/health")
async def health_check():
"""健康检查"""
return {"status": "ok", "service": "TTS Book Service", "api_key_configured": bool(config.MIMO_API_KEY)}
# ── 听书 App 音频接入接口 ─────────────────────────────────────────────────
@app.get("/api/book/{book_id}")
@@ -242,9 +349,6 @@ async def get_chapter_audio(book_id: str, chapter_id: str):
# ── 实时 TTS 接口（兼容听书 App 格式）─────────────────────────────────────
from fastapi.responses import Response
from urllib.parse import parse_qs
@app.post("/api/tts")
async def realtime_tts(request: Request):
"""
@@ -256,6 +360,7 @@ async def realtime_tts(request: Request):
"""
text = ""
style = ""
voice = ""
content_type = request.headers.get("content-type", "")
try:
@@ -263,13 +368,13 @@ async def realtime_tts(request: Request):
data = await request.json()
text = data.get("text", "").strip()
style = data.get("style", "").strip()
voice = data.get("voice", "").strip()
else:
# form-urlencoded (百度风格)
body_bytes = await request.body()
params = parse_qs(body_bytes.decode("utf-8"))
text = (params.get("tex", [""])[0]).strip()
# URL 解码（百度会 double-encode）
from urllib.parse import unquote
text = unquote(unquote(text))
except Exception:
pass
@@ -282,28 +387,60 @@ async def realtime_tts(request: Request):
)
try:
# MiMo TTS 生成 WAV
wav_bytes = await call_mimo_tts(text, style)
# 文本分段
chunks = split_text(text)
logger.info(f"实时 TTS: text_len={len(text)}, chunks={len(chunks)}, style={style or '(默认)'}, voice={voice or '(默认)'}")
# WAV → MP3（临时文件）
tmp_dir = Path(config.AUDIO_DIR) / "_tmp"
tmp_dir.mkdir(parents=True, exist_ok=True)
tmp_id = uuid.uuid4().hex
wav_path = str(tmp_dir / f"{tmp_id}.wav")
mp3_path = str(tmp_dir / f"{tmp_id}.mp3")
with open(wav_path, "wb") as f:
f.write(wav_bytes)
if len(chunks) == 1:
# 单段：直接生成
wav_bytes = await call_mimo_tts(text, style, voice)
tmp_id = uuid.uuid4().hex
wav_path = str(tmp_dir / f"{tmp_id}.wav")
mp3_path = str(tmp_dir / f"{tmp_id}.mp3")
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, mp3_path)
with open(wav_path, "wb") as f:
f.write(wav_bytes)
with open(mp3_path, "rb") as f:
mp3_bytes = f.read()
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, mp3_path)
# 清理临时文件
os.remove(wav_path)
os.remove(mp3_path)
with open(mp3_path, "rb") as f:
mp3_bytes = f.read()
os.remove(wav_path)
os.remove(mp3_path)
else:
# 多段：逐段生成 → 拼接
mp3_paths = []
for i, chunk in enumerate(chunks):
wav_bytes = await call_mimo_tts(chunk, style, voice)
chunk_id = uuid.uuid4().hex
wav_path = str(tmp_dir / f"{chunk_id}.wav")
mp3_path = str(tmp_dir / f"{chunk_id}.mp3")
with open(wav_path, "wb") as f:
f.write(wav_bytes)
loop = asyncio.get_event_loop()
await loop.run_in_executor(None, wav_to_mp3, wav_path, mp3_path)
os.remove(wav_path)
mp3_paths.append(mp3_path)
# 拼接所有 MP3
merged_id = uuid.uuid4().hex
merged_path = str(tmp_dir / f"{merged_id}.mp3")
await loop.run_in_executor(None, concat_mp3_files, mp3_paths, merged_path)
with open(merged_path, "rb") as f:
mp3_bytes = f.read()
# 清理
for p in mp3_paths:
os.remove(p)
os.remove(merged_path)
return Response(content=mp3_bytes, media_type="audio/mpeg")
@@ -459,7 +596,7 @@ async def generate_audio(book_id: str, chapter_id: str):
@app.post("/admin/api/books/{book_id}/generate-all")
async def generate_all_chapters(book_id: str):
"""批量生成书籍所有章节音频"""
"""批量生成书籍所有章节音频（并发，限制 3 路）"""
async with async_session() as db:
result = await db.execute(
select(Chapter).where(Chapter.book_id == book_id, Chapter.status != "ready")
@@ -467,10 +604,31 @@ async def generate_all_chapters(book_id: str):
chapters = result.scalars().all()
chapter_ids = [ch.chapter_id for ch in chapters]
for cid in chapter_ids:
await generate_chapter_audio(cid)
if not chapter_ids:
return {"ok": True, "total": 0, "chapter_ids": [], "message": "没有需要生成的章节"}
return {"ok": True, "total": len(chapter_ids), "chapter_ids": chapter_ids}
# 并发生成，限制同时 3 个请求避免过载
sem = asyncio.Semaphore(3)
async def _gen(cid: str):
async with sem:
await generate_chapter_audio(cid)
tasks = [_gen(cid) for cid in chapter_ids]
results = await asyncio.gather(*tasks, return_exceptions=True)
# 统计结果
errors = [str(r) for r in results if isinstance(r, Exception)]
success_count = len(chapter_ids) - len(errors)
return {
"ok": True,
"total": len(chapter_ids),
"success": success_count,
"failed": len(errors),
"errors": errors[:10] if errors else [],
"chapter_ids": chapter_ids,
}
# --- TTS 试听 ---
@@ -481,11 +639,12 @@ async def tts_preview(request: Request):
data = await request.json()
text = data.get("text", "").strip()
style = data.get("style", "").strip()
voice = data.get("voice", "").strip()
if not text:
raise HTTPException(400, "文本不能为空")
wav_bytes = await call_mimo_tts(text, style)
wav_bytes = await call_mimo_tts(text, style, voice)
audio_dir = Path(config.AUDIO_DIR) / "_preview"
audio_dir.mkdir(parents=True, exist_ok=True)
@@ -510,6 +669,7 @@ async def get_config():
"model": config.MIMO_TTS_MODEL,
"voice": config.MIMO_VOICE,
"api_key_masked": config.MIMO_API_KEY[:6] + "****" if config.MIMO_API_KEY else "未配置",
"max_chunk_chars": MAX_CHUNK_CHARS,
}