Skip to content

利用 AI 大模型和自动化工作流,根据主题或关键词一键生成高清短视频。Generate HD short videos from a topic or keyword with an automated AI workflow.

fix: harden Windows portable update and Azure TTS compatibility

Harry400c1c3f35be
diff --git a/app/services/voice.py b/app/services/voice.py
index 26ff8fcd..665b58b7 100644
--- a/app/services/voice.py
+++ b/app/services/voice.py
@@ -1,4 +1,5 @@
import asyncio
+import inspect
import math
import os
import re
@@ -1281,6 +1282,63 @@ def populate_legacy_submaker_with_full_text(
return sub_maker
+def create_edge_tts_communicate(
+ text: str, voice_name: str, rate_str: str
+) -> edge_tts.Communicate:
+ """
+ 按当前已安装的 edge_tts 版本构造 Communicate 对象。
+
+ 背景:
+ 1. 主线代码已经升级到 edge_tts 7.x,并使用 `boundary` 参数拿到更细的边界事件;
+ 2. 但 Windows 便携包如果更新失败,现场环境可能仍然停留在旧版 edge_tts;
+ 3. 旧版 `Communicate.__init__()` 不接受 `boundary`,会直接抛出
+ `unexpected keyword argument 'boundary'`,导致整个 TTS 链路失败。
+
+ 因此这里先根据构造函数签名探测当前版本支持的参数,再决定是否传入
+ `boundary`,让同一份代码同时兼容旧版和新版依赖。
+ """
+ communicate_kwargs = {"rate": rate_str}
+ communicate_signature = inspect.signature(edge_tts.Communicate)
+
+ if "boundary" in communicate_signature.parameters:
+ communicate_kwargs["boundary"] = "WordBoundary"
+
+ return edge_tts.Communicate(text, voice_name, **communicate_kwargs)
+
+
+def stream_edge_tts_chunks(communicate, on_chunk) -> None:
+ """
+ 统一消费 edge_tts 的同步流和旧版异步流。
+
+ edge_tts 7.x 提供 `stream_sync()`,可以在同步函数里直接迭代;
+ 更早的版本通常只有异步 `stream()`。为了让 `azure_tts_v1()` 在
+ 旧依赖残留场景下仍能继续工作,这里统一做一层流式兼容。
+
+ Args:
+ communicate: edge_tts.Communicate 实例
+ on_chunk: 每拿到一个事件块时执行的回调
+ """
+ if hasattr(communicate, "stream_sync"):
+ for chunk in communicate.stream_sync():
+ on_chunk(chunk)
+ return
+
+ if not hasattr(communicate, "stream"):
+ raise AttributeError("edge_tts communicate object has no stream method")
+
+ async def _consume_async_stream():
+ async for chunk in communicate.stream():
+ on_chunk(chunk)
+
+ # 这里显式创建独立事件循环,而不是复用外部上下文,目的是避免
+ # 在同步调用栈里遇到“当前线程没有事件循环”或跨线程复用循环的问题。
+ loop = asyncio.new_event_loop()
+ try:
+ loop.run_until_complete(_consume_async_stream())
+ finally:
+ loop.close()
+
+
def azure_tts_v1(
text: str, voice_name: str, voice_rate: float, voice_file: str
) -> Union[SubMaker, None]:
@@ -1291,27 +1349,26 @@ def azure_tts_v1(
try:
logger.info(f"start, voice name: {voice_name}, try: {i + 1}")
- # edge_tts 7.x 已经提供同步流接口和新的字幕聚合器实现,
- # 这里直接使用同步接口,避免继续依赖旧版本的异步流事件结构。
+ # 这里同时兼容 edge_tts 7.x 和旧版便携包里可能残留的老依赖:
+ # 1. 新版支持 `boundary` + `stream_sync()`
+ # 2. 旧版不支持 `boundary`,且通常只暴露异步 `stream()`
ensure_file_path_exists(voice_file)
- communicate = edge_tts.Communicate(
- text,
- voice_name,
- rate=rate_str,
- boundary="WordBoundary",
- )
+ communicate = create_edge_tts_communicate(text, voice_name, rate_str)
sub_maker = edge_tts.SubMaker()
with open(voice_file, "wb") as file:
- for chunk in communicate.stream_sync():
+ def _handle_chunk(chunk):
chunk_type = chunk["type"]
if chunk_type == "audio":
file.write(chunk["data"])
elif chunk_type in ["WordBoundary", "SentenceBoundary"]:
- # 直接把 7.x 返回的边界事件喂给新的 SubMaker,
- # 这样后续可以复用官方生成的字幕与时间轴。
+ # 无论来自 7.x 的同步流,还是旧版异步流,只要事件结构
+ # 里仍有边界信息,就统一喂给 SubMaker,保证后续字幕链路
+ # 仍然走项目现有逻辑。
sub_maker.feed(chunk)
+ stream_edge_tts_chunks(communicate, _handle_chunk)
+
if not sub_maker.get_srt():
logger.warning("failed, sub_maker.get_srt() is empty")
continue
diff --git a/requirements.txt b/requirements.txt
index e8ab37b2..cc8a63c5 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,4 +1,4 @@
-# 兼容旧的 pip 安装方式;主依赖定义已迁移到 pyproject.toml / uv.lock。
+# Keep legacy pip install support; primary deps moved to pyproject.toml / uv.lock.
moviepy==2.1.2
streamlit==1.45.0
edge_tts==7.2.7
diff --git a/test/services/test_voice.py b/test/services/test_voice.py
index 4f2f175e..e02ee7e4 100644
--- a/test/services/test_voice.py
+++ b/test/services/test_voice.py
@@ -86,6 +86,59 @@ class TestVoiceService(unittest.TestCase):
audio_duration = vs.get_audio_duration(sub_maker)
print(f"voice: {voice_name}, audio duration: {audio_duration}s")
+ def test_azure_tts_v1_supports_legacy_edge_tts_without_boundary(self):
+ """
+ 验证 Azure TTS V1 在旧版 edge_tts 依赖残留时仍可继续工作。
+
+ 这个回归场景对应 Windows 便携包更新失败后,现场环境还停留在旧版
+ edge_tts 的情况:
+ 1. `Communicate.__init__()` 不接受 `boundary`
+ 2. 只有异步 `stream()`,没有 `stream_sync()`
+ """
+
+ class _LegacyCommunicate:
+ def __init__(self, text, voice, rate="+0%"):
+ self.text = text
+ self.voice = voice
+ self.rate = rate
+
+ async def stream(self):
+ yield {"type": "audio", "data": b"legacy-audio"}
+ yield {
+ "type": "WordBoundary",
+ "offset": 0,
+ "duration": 10000000,
+ "text": "legacy",
+ }
+
+ class _FakeSubMaker:
+ def __init__(self):
+ self.events = []
+
+ def feed(self, chunk):
+ self.events.append(chunk)
+
+ def get_srt(self):
+ if not self.events:
+ return ""
+ return "1\n00:00:00,000 --> 00:00:01,000\nlegacy\n"
+
+ with tempfile.TemporaryDirectory() as tmp_dir, patch.object(
+ vs.edge_tts, "Communicate", _LegacyCommunicate
+ ), patch.object(vs.edge_tts, "SubMaker", _FakeSubMaker):
+ voice_file = str(Path(tmp_dir) / "legacy-edge-tts.mp3")
+ sub_maker = vs.azure_tts_v1(
+ text="legacy edge tts compatibility",
+ voice_name="zh-CN-XiaoyiNeural-Female",
+ voice_file=voice_file,
+ voice_rate=1.0,
+ )
+
+ self.assertIsNotNone(sub_maker)
+ self.assertEqual(Path(voice_file).read_bytes(), b"legacy-audio")
+ self.assertEqual(len(sub_maker.events), 1)
+ self.assertEqual(sub_maker.events[0]["type"], "WordBoundary")
+
def test_azure_tts_v2(self):
voice_name = "zh-CN-XiaoxiaoMultilingualNeural-V2-Female"
voice_name = vs.parse_voice_name(voice_name)