Product
Image-to-video workflow for Seedance 2.5 AI
Reference images are strongest when the prompt explains motion, not just appearance. This workflow turns still frames into controlled moving shots.
InputImage + promptUse a clear reference frame
ControlMotion-firstPrompt what should change
OutputReusable clipDownload or iterate from history
For image-to-video, the still frame already carries subject, composition, and style. The text prompt should focus on movement: camera drift, environmental motion, product rotation, character gesture, or lighting change.
Keep the reference image readable
Use images with one dominant subject, stable framing, and enough negative space for motion. If the source image is crowded, the model has less room to interpret movement cleanly.
Prompt pattern
Preserve the subject and composition from the reference image. Add a slow camera move, a single subject action, and a clear lighting or atmosphere change.