Street encounter
Wide · eye level · 28mm · crane down
She walks towards camera, he walks away; they meet at centre frame.
Flowith 3.0 · 3D Director Stage
A 3D stage inside your canvas. Block the scene with figures, choose the lens and the move, and the AI video follows your camera, not its own guess.
Try it here
Change the lens, the angle and the move. The shot language updates as you go, exactly how the real stage writes the prompt.
A sketch of the real thing. On the canvas the stage has 19 poses, 21 set pieces, 15 camera moves, a keyframe timeline and as many cameras as you need.
Three stage templates, shot by Seedance 2.5 at 1080p from the grey previs. The only thing we added was one line about the look.
Wide · eye level · 28mm · crane down
She walks towards camera, he walks away; they meet at centre frame.
Medium · eye level · 35mm · truck left
He runs screen left; the camera tracks with him.
Full shot · eye level · 35mm · slow push in
She talks with her hands; he listens.
Men, women and children on the floor. 19 poses from walk and run to sit, aim and hug, or pose every joint by hand.
Walls, doors, tables, cars, trees and streetlights in real-world sizes. Turn a picture into a 3D prop.
Studio, room or street, or a 360° backdrop painted from words. Six light presets from golden hour to noir.
14 to 135mm, handheld shake, and 15 one-click moves: push, pull, pan, orbit, crane, dolly zoom, arc.
Move a figure at a later moment and it walks there. Keyframes for every figure and camera.
Seven classic stagings, or describe the scene in a sentence and AI lays out the figures, set and cameras.
Shot size, angle, lens, move and who stands where, kept in sync with the stage and sent with the previs.
Render the grey frame into a real picture with the same framing, or export depth, normal and pose passes.
Start from a template, describe it to AI, or drag figures onto the floor yourself.
Switch to camera view, choose the lens and the move, add as many cameras as the scene needs.
The stage records the previs, writes the prompt and opens a Seedance generator on your canvas. Check it, then send.
Blocking, camera moves and the previs recording run in your browser. Credits are only used when you generate: the shot itself, a scene built by AI, a 360° backdrop or a 3D prop.
The best tutorials we found on blocking a shot in 3D and handing it to an AI video model, plus the directing basics behind every good shot.
Videos and posts belong to their creators and open from YouTube and X.
A card on the Flowith canvas that opens a 3D stage. You block a scene with figures, props and cameras, then shoot it: the stage records a grey previs and hands it to an AI video model together with the written shot language, so the video follows your camera and blocking.
No. Start from one of seven templates or describe the scene in a sentence and let AI build it. Figures drag around on the floor, and a camera can be added from whatever you are looking at.
Seedance 2.5 by default, or the Seedance 2.x model you already picked. The previs goes in as a video reference, so the camera move, framing and positions carry over while the look is up to your prompt.
No. Everything on the stage runs locally. Credits are spent only when you generate the video, a still, an AI-built scene, a 360° backdrop or a 3D prop.
On any canvas in Flowith 3.0: add a card and pick 3D Director Stage. From a storyboard shot, choose Block this shot in 3D and the clip goes back into that shot when it is ready.
Open a canvas, add a 3D Director Stage and block your first scene.
Start directing