静帧 + 配音 → Seedance，换人设不重剪
2.5 出首帧、配音另做，丢进 Seedance 2.5 成片后再换角色——同一套剪辑不用重来。短视频人设 A/B 测的正经流水线。
Demo video
No result image in the original post
Prompt
ai is getting uncomfortably real. both of these are ai-generated. same script, same edit, different people. we made the first frame(gpt image 2.5) and voice separately, then gave both to seedance 2.5. once the base video worked, we changed the character without rebuilding the edit. getting the face, voice and timing right took a few attempts. but look at the results dm me if you want full workflow pdf
Attribution
- Author
- meetshukla_
- Published
- 2026-09-11
- Original post
- View original ↗
- Collected from
- xianyu110/awesome-gpt-image2.5 · 2098280866459861176 · MIT