From 22e04d65a76971f86ad16f96cbf45fd1c603f6a8 Mon Sep 17 00:00:00 2001
From: xiaoxia <xiaoxia@xiaoxiajianji.com>
Date: Wed, 30 Sep 2026 20:42:12 +0800
Subject: [PATCH] =?UTF-8?q?refactor(viral-video):=20#2106=20=E5=88=A0?=
 =?UTF-8?q?=E9=99=A4=20=5Fstep=5Fmusetalk=20=E6=AD=A5=E9=AA=A4?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

爆款视频由 Seedance 2.5 直接生成人物口型，不需要 MuseTalk 事后对口型。
MuseTalk 是 AI 数字人路线（上传人物视频+配音→对嘴型）用的，跟爆款视频是两条不同路线。

- 删除 _step_musetalk 函数
- resume_pipeline 直接把 render 输出传给 upload
- 更新流水线 docstring 为 9 步
- 删除/更新对应单测
- ViralVideoStage.MUSETALK 枚举值保留以避免前端 breaking change
---
 apps/worker/worker_app/tasks/viral_video.py | 27 ++++++---------------
 tests/unit/test_viral_video.py              |  3 ---
 tests/unit/test_viral_video_p0.py           | 15 ------------
 3 files changed, 7 insertions(+), 38 deletions(-)

diff --git a/apps/worker/worker_app/tasks/viral_video.py b/apps/worker/worker_app/tasks/viral_video.py
index 6ebedcb07..835ed0683 100644
--- a/apps/worker/worker_app/tasks/viral_video.py
+++ b/apps/worker/worker_app/tasks/viral_video.py
@@ -1,6 +1,6 @@
 """爆款视频 Celery 编排器 — ViralVideoOrchestrator.
 
-10 步流水线：
+9 步流水线（Seedance 2.5 直生口型，不再走 MuseTalk）：
   1. 图片 VLM 分析
   1.5 [v1.3] 视频风格分析（如用户上传参考视频）
   2. 用户文案意图解析
@@ -8,10 +8,9 @@
   4. 分镜脚本生成
   5. 合规审核（6 维度，不通过自动重写 1 次）
   6. CosyVoice 配音
-  7. BGM 选择
-  8. UnifiedRenderService 渲染
-  9. 数字人口型（MuseTalk）
-  10. OSS 上传 + 通知 + 扣点
+  7. BGM 选择（素材未就绪时跳过）
+  8. Seedance 逐分镜生成 + ffmpeg concat + 混 TTS
+  9. OSS 上传 + 通知 + 扣点
 """
 
 from __future__ import annotations
@@ -593,14 +592,6 @@ def _make_placeholder_clip(tmpdir, idx: int, duration: int):
     return out
 
 
-def _step_musetalk(job: ViralVideoJob, video_path: str) -> str:
-    """步骤 9: 数字人口型。P1：MuseTalk GPU 端点未就绪前统一跳过。"""
-    if not job.persona_id:
-        return video_path
-    logger.info("[爆款视频] persona_id=%s 已设置，但 MuseTalk GPU 端点未就绪，跳过口型同步", job.persona_id)
-    return video_path
-
-
 def _step_upload(job: ViralVideoJob, video_path: str) -> str:
     """步骤 10: OSS 上传。"""
     try:
@@ -760,14 +751,10 @@ def resume_viral_video_pipeline(self: Task, job_id: str) -> dict:
         video_path = _step_render(job, storyboard, tts_path, bgm)
         _emit_progress(job_id, ViralVideoStage.RENDERING, 88.0, "渲染完成")
 
-        # ── Step 9: MuseTalk（P1：端点未就绪前跳过） ──
-        _emit_progress(job_id, ViralVideoStage.MUSETALK, 90.0, "正在处理数字人口型...")
-        final_video_path = _step_musetalk(job, video_path)
-        _emit_progress(job_id, ViralVideoStage.MUSETALK, 93.0, "数字人处理完成")
-
-        # ── Step 10: OSS 上传 + 扣点 ──
+        # ── Step 9: OSS 上传 + 扣点 ──
+        # 注：爆款视频由 Seedance 2.5 直接生成口型，不需要 MuseTalk 事后对口型（MuseTalk 是 AI 数字人路线用的）。
         _emit_progress(job_id, ViralVideoStage.UPLOADING, 95.0, "正在上传视频...")
-        video_url = _step_upload(job, final_video_path)
+        video_url = _step_upload(job, video_path)
 
         job.credits_cost = CREDITS_VIRAL_VIDEO_COST
         # TODO: 调用 credits.deduct() 实际扣点（#1895 总开关为 false 时不扣，保留 TODO）
diff --git a/tests/unit/test_viral_video.py b/tests/unit/test_viral_video.py
index 73cc13f43..a18bf35b7 100755
--- a/tests/unit/test_viral_video.py
+++ b/tests/unit/test_viral_video.py
@@ -456,7 +456,6 @@ class TestPipelineIntegration:
     """流水线端到端集成测试（mock 外部依赖）。"""
 
     @patch("apps.worker.worker_app.tasks.viral_video._step_upload")
-    @patch("apps.worker.worker_app.tasks.viral_video._step_musetalk")
     @patch("apps.worker.worker_app.tasks.viral_video._step_render")
     @patch("apps.worker.worker_app.tasks.viral_video._step_bgm_select")
     @patch("apps.worker.worker_app.tasks.viral_video._step_tts")
@@ -481,7 +480,6 @@ class TestPipelineIntegration:
         mock_tts,
         mock_bgm,
         mock_render,
-        mock_musetalk,
         mock_upload,
     ):
         """测试 resume 流水线能从确认状态走到完成。"""
@@ -509,7 +507,6 @@ class TestPipelineIntegration:
         mock_tts.return_value = None  # P1: TTS 返回 Path|None，mock 用 None 跳过混音
         mock_bgm.return_value = None  # P1: BGM 未就绪前返回 None
         mock_render.return_value = "/tmp/video.mp4"
-        mock_musetalk.return_value = "/tmp/video_final.mp4"
         mock_upload.return_value = "https://oss.example.com/final.mp4"
 
         result = resume_viral_video_pipeline.run("job-001")
diff --git a/tests/unit/test_viral_video_p0.py b/tests/unit/test_viral_video_p0.py
index 7417b9d84..d5d98b8c4 100644
--- a/tests/unit/test_viral_video_p0.py
+++ b/tests/unit/test_viral_video_p0.py
@@ -155,21 +155,6 @@ class TestBGMSkip:
         assert _step_bgm_select(mock_job) is None
 
 
-class TestMuseTalkSkip:
-    def test_musetalk_no_persona_passes_through(self, mock_job):
-        from apps.worker.worker_app.tasks.viral_video import _step_musetalk
-
-        mock_job.persona_id = ""
-        assert _step_musetalk(mock_job, "/tmp/video.mp4") == "/tmp/video.mp4"
-
-    def test_musetalk_with_persona_still_skips(self, mock_job):
-        """GPU 端点未就绪前，即使有 persona 也直接返回原路径。"""
-        from apps.worker.worker_app.tasks.viral_video import _step_musetalk
-
-        mock_job.persona_id = "persona-1"
-        assert _step_musetalk(mock_job, "/tmp/video.mp4") == "/tmp/video.mp4"
-
-
 # ── P0-1: call_video_generation 参数构造 ──────────────────────────
 
 
-- 
2.54.0