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.
Python ai-video-generatorcontent-creationffmpeginstagram-reelsllmpythonshort-videosubtitlestext-to-speechtiktokvideo-automationvideo-workflowworkflow-automationyoutube-shorts
docs: add PR merge and validation record for 2026-04-02
Harry9db81e728c59
diff --git a/docs/PR_MERGE_RECORD_2026-04-02.md b/docs/PR_MERGE_RECORD_2026-04-02.mdnew file mode 100644index 00000000..6b6a8cd8--- /dev/null+++ b/docs/PR_MERGE_RECORD_2026-04-02.md@@ -0,0 +1,120 @@+# 2026-04-02 PR 合并与验证记录++## 本次已合并并推送的 PR++- `#837` `fix: update google-generativeai version for response_modalities support`+- `#835` `fix: add missing pydub dependency to requirements.txt`+- `#850` `feat: support reading subtitle position from config file`+- `#838` `feat: add MiniMax as LLM provider`+- `#811` `refactor: optimize codebase for better performance and reliability`+- `#848` `feat: support GPU acceleration for faster-whisper in Docker`+- `#843` `feat: Add Upload-Post integration for cross-posting to TikTok/Instagram`++## 合并后的主线提交++- TTS 与字幕修复基线提交:`953a6c0` `fix: restore edge tts synthesis and readable subtitles`+- 当前主线提交:`1f8a746`++## 合并时的验证结论++### 已通过++- `#837`+ - 依赖升级后可正常导入+ - `google-generativeai==0.8.6` 已生效+- `#835`+ - `pydub==0.25.1` 已生效+- `#850`+ - `subtitle_position` 与 `custom_position` 可从配置文件读取+- `#838`+ - MiniMax provider 接线正常+ - 使用 mock 调用验证 `_generate_response` 通过+- `#811`+ - 主线导入正常+ - 抽样单测通过+- `#848`+ - `docker compose -f docker-compose.yml -f docker-compose.gpu.yml config` 可正常解析+- `#843`+ - Upload-Post 服务导入和 mock 上传调用通过+ - 与前面 PR 叠加时仅在 `config.example.toml` 存在配置段落冲突,已手工保留两边内容++### 已拒绝并关闭++- `#852`+ - 能恢复音频,但会破坏字幕链路,并删除仍被 WebUI 调用的 Gemini 逻辑+- `#787`+ - 不能解决当前 `403` 场景+- `#841`+ - 与当前主线 TTS/字幕修复冲突,且收益已被更小 PR 覆盖+- `#824`+ - ModelsLab 路径能出音频,但字幕链路失败,无法产出可用 SRT+- `#840`+ - 后端加入 `video_source="ai"`,但 WebUI 仍不支持该值,端到端不可用+- `#826`+ - 与当前主线 `voice.py` 和依赖变更冲突,未通过合并验证+- `#751`+- `#749`+- `#742`+- `#705`+ - 以上 4 个 PR 在当前主线下均为 `DIRTY`,未通过合并验证++## 冒烟测试记录++### 服务重启++- API:`http://127.0.0.1:8080/docs`+- WebUI:`http://127.0.0.1:8501`++### 第一次完整视频任务++- 任务号:`ced0b190-dd72-489c-b978-2761740933db`+- 结果:失败+- 结论:+ - API 默认 `video_transition_mode=null`+ - 视频拼接阶段在 `app/services/video.py` 中直接访问 `video_transition_mode.value`+ - 导致任务线程异常退出,任务状态停留在 `state=4, progress=75`++### 第二次完整视频任务++- 任务号:`8b2a0e6e-b3e6-44ab-a1b4-1865a0b4788d`+- 提交方式:+ - `POST /api/v1/videos`+ - 使用本地素材 `/Users/harry/Projects/Python/MoneyPrinterTurbo/test/resources/1.png`+ - 显式指定 `video_transition_mode="FadeIn"`+- 结果:成功+- 任务状态:`state=1, progress=100`++### 第二次任务产物++- 音频:`/Users/harry/Projects/Python/MoneyPrinterTurbo/storage/tasks/8b2a0e6e-b3e6-44ab-a1b4-1865a0b4788d/audio.mp3`+ - 时长:`8.952s`+ - 大小:`53712 bytes`+- 拼接视频:`/Users/harry/Projects/Python/MoneyPrinterTurbo/storage/tasks/8b2a0e6e-b3e6-44ab-a1b4-1865a0b4788d/combined-1.mp4`+ - 时长:`9.000s`+ - 大小:`177666 bytes`+- 成片:`/Users/harry/Projects/Python/MoneyPrinterTurbo/storage/tasks/8b2a0e6e-b3e6-44ab-a1b4-1865a0b4788d/final-1.mp4`+ - 时长:`9.000s`+ - 大小:`352810 bytes`+- 字幕:`/Users/harry/Projects/Python/MoneyPrinterTurbo/storage/tasks/8b2a0e6e-b3e6-44ab-a1b4-1865a0b4788d/subtitle.srt`++### 第二次任务字幕样本++```srt+1+00:00:00,100 --> 00:00:03,300+这是一次主线合并后的完整冒烟测试++2+00:00:03,875 --> 00:00:05,350+我们要确认语音++3+00:00:05,575 --> 00:00:08,375+字幕和视频成片都能正常生成+```++## 当前仍需关注的风险++- `#843` 仅做了 mock 验证,尚未使用真实 Upload-Post 密钥联调+- `#848` 仅验证了 Docker GPU 配置解析,尚未在真实 GPU 环境运行+- 当前 API 默认 `video_transition_mode=null` 时,完整视频任务仍存在回归风险