SPICYFRAME · PRACTICAL GUIDES

How to use multiple reference images for AI video

Choose consistent reference photos, label their roles and build a clear multi-image video prompt in SpicyFrame.

Give every photo a job

Multiple references are most useful when they explain different parts of the same scene. Use one clear image for the subject, another for the setting and, if needed, another for clothing or lighting. SpicyFrame accepts up to 12 references, but filling every slot is not a requirement. Two compatible images can give clearer direction than twelve conflicting ones. Only upload images you own or have permission to use.

Choose consistent source material

Compare your references before uploading. Large changes in hairstyle, outfit, age, lighting or camera angle can make the desired result ambiguous. Remove near-duplicates and references that contradict your main photo. If preserving an outfit matters, show it clearly in the primary reference instead of expecting the model to reconstruct details hidden outside the frame.

Tell the prompt what to borrow

Try: “Use @Image 1 as the subject and outfit reference. Use @Image 2 for the room and warm window lighting. The subject turns slowly toward the window. One continuous medium shot, gentle camera push-in, natural room ambience.” The numbered labels follow the uploaded references; check their order before payment. The request guides the model rather than guaranteeing exact identity or pixel-for-pixel reproduction.

Change one thing per attempt

Start with a short Medium render to evaluate whether the references work together. If the background is wrong, simplify the setting instruction before changing the camera and action too. Comparing one adjustment at a time makes it easier to understand what helped. A new render is a separate paid request, so check the displayed piment cost before submitting.

When one reference is enough

For a subtle portrait movement or a slow camera move around an existing composition, a single clean source often conveys the whole scene. Add another image only when it supplies information that the first image lacks. Review faces, hands and clothing in the completed video before downloading or publishing.

Create your next scene

Keep exploring