Upload one or two reference images of a character, mascot, or product, describe the scene, and get a clip where that subject looks the same as in your references.
From one or two pictures of your subject to a clip of it in a new scene, in three steps.
A text prompt alone invents a new person every time. A reference image tells the AI who or what has to be in the shot: your face, a mascot, a character design, a product. You describe the scene, and the clip is built around that subject. That is how the same artist ends up in several short clips for a teaser campaign.
Only for a very short clip. A new web account starts with 20 credits, enough for one Standard clip of 2 seconds at 720p (19 credits) or 3 seconds at 360p (18 credits). The default Standard clip, 5 seconds at 720p, costs 47 credits, and Best costs 12 credits a second, from 60 for 5 seconds. The price shows on the Generate button, and credits are refunded if a generation fails.
It takes two reference images at most, and only the Standard tier uses the second. The Best tier works from the first image alone and starts the clip on that image, so there it behaves like Image to Video. Likeness is not guaranteed: faces, logos, and small text can change. A clip runs 15 seconds at most, and a video cannot be used as a reference. For a whole music video with the same cast in every shot, use the Solmi AI music video generator.
Image to video animates one picture as the first frame. Reference to video, on its Standard tier, uses your images as a guide to the subject and generates a new scene around it.
One or two. One clear image works, and the Standard tier also uses a second, such as another angle. The Best tier uses only the first image, as the opening frame of the clip.
For a full video, Solmi's AI music video generator keeps a cast consistent across every shot automatically.
Video generation guided by images of the subject. You upload one or two reference images and describe a scene, and the AI makes a clip in which that person, character, or product appears.
No. This tool takes still images as references. To make a photo follow the movement in a video, use Motion Transfer, which takes a character photo and a motion video.