What helps keep a reference image consistent in AI video?
Hi everyone! I work on Joyou AI, a creative workspace for AI image generation, video creation, and text chat. I'm interested in how other makers handle the jump from a still image to a short clip.
A typical workflow is to draft the concept and prompt in chat, generate or edit a still, and use it as the video's reference image. For a night scene, the still might have warm storefront light and cool street shadows. In the video, I'd want that lighting and the storefront layout to stay recognisable while the camera moves.
If you've tried this, what tends to work better for consistency: keeping the motion prompt short, specifying what must stay fixed, or using a different reference frame?
Joyou AI's English workspace, for context: joyou-ai.com/workbench/home?lng=en




