Image
Reference image
A reference image is a picture supplied alongside a prompt so the model holds something from it — a face, a product, a style — across new renders. It is the basis of character consistency in AI image and video.
For faces, a few clear photos hold identity through new poses, lighting and wardrobe. For products, a reference keeps the same object in every shot. In video, the first reference becomes the start frame of an image-to-video clip.
On eroq
Up to four reference pictures per image render (ten on Eroq One), added with + Picture; a Character's photos are references automatically when @mentioned. In video, a picture you attach becomes the start frame; without one, the cast and elements ride as numbered reference pictures — up to six on Seedance, four on Kling, three on Veo 3.1, Veo 3.1 Fast, Hailuo 3 and Wan 3.0.
Questions
How many reference photos should I use?
Two to four clear, well-lit photos of the same face are plenty.
Do references cost extra credits?
On images, no — reference photos ride along free (up to four, ten on Eroq One), a Character's or Element's included. On video, only Hailuo 3 bills them, at 17 credits a picture; every other engine reads its reference pictures at the normal render price.
Do references work on video too?
Yes, two ways. A picture you attach becomes the start frame and the engine animates it — on every engine — so put the opening shot first. Without a start frame, your cast and elements ride as numbered reference pictures on the engines that read them: up to 6 on Seedance, 4 on Kling, 3 on Veo 3.1, Veo 3.1 Fast, Hailuo 3 and Wan 3.0.
More image terms