auto-save 2026-05-14 10:25 (~13)

api/.env.example

@@ -14,11 +14,11 @@ VIDEO_MODEL_VEO3=veo-3.1-fast
 
 # 音频文案改写 + MiniMax 配音
 AUDIO_REWRITE_MODEL=gemini-2.5-pro
-AUDIO_PRODUCT_BRIEF=SKG 智能按摩产品，主打日常肩颈、腰背、眼部、膝盖或足部放松；广告表达要高级、干净、可信，不做医疗疗效承诺。
+AUDIO_PRODUCT_BRIEF="SKG 智能按摩产品，主打日常肩颈、腰背、眼部、膝盖或足部放松；广告表达要高级、干净、可信，不做医疗疗效承诺。"
 MINIMAX_API_KEY=
 MINIMAX_TTS_BASE_URL=https://api.minimax.io
 MINIMAX_TTS_MODEL=speech-2.8-turbo
-MINIMAX_TTS_VOICE_ID=Chinese (Mandarin)_Reliable_Executive
+MINIMAX_TTS_VOICE_ID="Chinese (Mandarin)_Reliable_Executive"
 
 # Poe 视频 API（优先用于 Seedance / Kling / Veo）
 POE_API_BASE_URL=https://api.poe.com/v1
@@ -41,6 +41,7 @@ VIDEO_CONTENT_PATH=/videos/{id}/content
 VIDEO_DURATION_FIELD=seconds
 
 # 工作目录
+KEYFRAME_COUNT=12
 JOBS_DIR=./jobs
 
 # CORS

api/README.md

@@ -9,7 +9,7 @@ cd api
 python3 -m venv .venv
 source .venv/bin/activate
 pip install -r requirements.txt
-cp .env.example .env  # 按需填 GEMINI_API_KEY
+cp .env.example .env  # 按需填 LLM_API_KEY / MINIMAX_API_KEY
 uvicorn main:app --port 4291 --reload
 ```
 

api/main.py

@@ -104,7 +104,7 @@ JobStatus = Literal[
     "transcribing", "transcribed", "failed",
 ]
 
-KEYFRAME_COUNT = int(os.getenv("KEYFRAME_COUNT", "5"))
+KEYFRAME_COUNT = int(os.getenv("KEYFRAME_COUNT", "12"))
 FrameExtractTarget = Literal["transparent_human", "balanced", "subject", "transition", "expression", "motion"]
 FrameExtractMode = Literal["replace", "append"]
 FrameExtractQuality = Literal["auto", "fast", "accurate", "ultra"]
@@ -1327,7 +1327,7 @@ async def analyze_queue_worker() -> None:
         ANALYZE_WORKER_RUNNING = False
 
 
-# ---------- Gemini ASR + 翻译 ----------
+# ---------- 音频转写 + 翻译 + SKG 改写 + MiniMax 配音 ----------
 
 def _transcribe_sync(wav: Path) -> list[dict]:
     """whisper-1 verbose_json → segments[{start, end, text}]"""
@@ -1347,7 +1347,7 @@ def _transcribe_sync(wav: Path) -> list[dict]:
 
 
 def _translate_sync(segments: list[dict]) -> list[str]:
-    """gemini-2.5-flash 批量翻译为中文，按段返回"""
+    """批量翻译为中文，按段返回"""
     payload = [{"i": i, "en": s.get("text", "").strip()} for i, s in enumerate(segments)]
     prompt = (
         "你是字幕翻译。把下列英文字幕段翻译为简体中文，保持原意、口语化、自然流畅。"