Lock the references
Front, three-quarter, full body. Same outfit.
角色一致 / 同一张脸
Give the model one to several reference images of a character — face, wardrobe, the room — and describe each shot. Identity is carried across takes, so a scene can be built from many clips instead of one lucky one.
最多七张参考:脸、衣服、房间。从进门拍到结束,还是同一个人。
How it works
Front, three-quarter, full body. Same outfit.
One motion per clip; keep the setting constant.
Generate each shot, then extend or cut together.
You give
You get
Under the hood
Each model lists its own price and content policy. The spicy route is MiniMax H3; other models follow their vendor’s rules.
FAQ
Three is the sweet spot: face close-up, three-quarter, full body. More references help wardrobe; too many contradictory ones hurt.
Only an adult who has consented. Synthetic characters are the intended use.
A little on 15 s takes. Keep clips short and use the same references for every shot.
Enter an invite code to open the tool in Chat, or join the waitlist from this page. We open seats in batches and prioritise the tools people ask for most.
Related tools
Invite code opens the tool. No code — join the waitlist; the most-requested tools open first.
Early access
Consistent character video is invite-only for now. Enter a code, or leave your email and we will count the request.