Every film on Visuall is a canvas of boxes. Each box does one job: write the scenes, make the images, animate them, join the clips. Wire them together, approve what you like on the way, and change any step later.
The agent builds your canvas for free.
Here is one real canvas for an AI influencer: a made-up travel creator, four faces to choose from, posts in Lisbon and a short reel.
Say what you want in your own words, the way you would brief a person, and attach photos if you have them. The canvas agent builds the canvas for it: every box, model and wire. Building a new canvas is free. Later, ask for changes in the bar under the canvas, 2 credits a change. The agent only changes the canvas: nothing is made or charged until you press Run.
A box does one thing: hold your text, make four faces with FLUX, put the chosen face into each post with Nano Banana, turn a photo into a clip. Wires carry each box's results to the next, so you can always see where an image came from. Add boxes from the library on the left, or drag one onto the canvas.
Try this example →
Make four faces and keep one. Make four posts and approve three. Only what you approve in a Pick box moves on, and with "Wait for my pick" the run stops there until you choose. You never pay to animate a frame you would throw away.
Its settings open on the right: the model and what it costs, the length, sound and quality. Run to here makes just that box and the boxes it needs. Regenerate makes it again for a fresh take.
Each box shows its price before you run. A run sets aside its estimate and gives back what it did not use, and failed results are not charged. Change one box and run again: everything that did not change is reused for free. At the end, Stitch Videos joins the clips and Captions writes your words on them.
Grouped the way the box library groups them. Image and video boxes let you choose their model.
Each is a canvas the agent wires when you describe that kind of job. Send the example as it is or write your own, then open any box and change it: the model, how many variations, what gets approved.
Story in, finished video out. The AI writes timed scenes, Gemini Omni films each one, and the clips are stitched into one video you can caption scene by scene. Add a style and it carries into every scene.
Try this example →The way Filmmaker worked: scenes become keyframes you approve, then each approved frame is animated with its scene's length, stitched and captioned. Your style is added to every keyframe prompt.
Try this example →Your story as 3D scenes in one look, grey clay by default. You approve the keyframes before they are animated.
Try this example →One product photo placed into the scenes you list, keeping its shape, label and colours. Approve the shots, and they are animated into a short ad with your captions.
Try this example →A consistent persona: pick a face from four, put that same person into post photos, then animate the posts you approve into a reel and caption it. A body photo and outfits are optional extras. The run waits for each pick.
Try this example →Start from a real face instead of a made-up one. Add a body photo and outfits if you have them: the posts keep that body and put the person in those clothes. Approve the posts, and they become a captioned reel.
Try this example →Each image and video box lets you choose its model, so one workflow can mix them.