A 3D scene tells a video model what a prompt cannot: where each person stands, which way they face, and how the camera moves. This guide takes a blocked 3D scene to a finished AI video in Morphic, in six steps.

Finished rainy night shot from the same low camera: a courier in a yellow rain jacket faces a tall man in a dark coat standing at a lit doorway in a brick alley

Generated shot

3D Scene blocking

What you need to turn a 3D scene into video

No modelling skills needed, just three things. The scene can come from Morphic's 3D Scene or any 3D tool.

A 3D scene of the shotRendered as a blocking video or a still.
A line per figureWho each figure is and what they wear, cast by colour.
A reference sheet eachFor any character whose face and costume must hold.

The 3D-to-video workflow

Describe the shot and let Copilot block it

Describe the shot in 3D Scene: a rainy alley, a courier backing away, a man in a doorway, a low camera pushing in. Copilot builds it, or you place it by hand in the 3D scene generator.

Refine by hand and clear the continuity warnings

Move anyone, redraw a path or swap the lens, then play the timeline. Fix any 180-degree line crossing or odd speed the editor flags, while it costs nothing.

Render the blocking video and add it to the Canvas

Render a Video from Export at 720p or 1080p, in your browser at no credit cost. Choose Add video to canvas, or press P for a snapshot when one frame is enough.

Prompt by figure colour

Attach the blocking video and give each figure one line by colour. The colours tell the model who goes where; the blocking guides which way each one faces.

Prompt by figure colour

The red figure is a courier in a yellow rain jacket. The blue figure is a tall man in a dark wool coat. Keep the camera, the staging and the push-in from the reference.

Editar prompt

Generate with reference sheets for look and identity

Add each character's reference sheet, pick a video model that accepts a reference video, and generate two versions. Keep the one closest to the blocking.

Assemble the shots in Compose

Repeat for each camera, then cut the shots together in Compose: trim, order, grade and add sound. Every shot starts from the same geography, so the cuts line up.

Common problems when turning a 3D scene into video

Most problems trace back to the prompt or to one camera doing too much. Planning a whole sequence? See previs software. From Blender, follow the Blender render to AI video guide, and browse 3D animation videos for finished looks.

Problem	Cause	Fix
Wrong person in the wrong place	No colours in the prompt	One line per figure: "The red figure is..."
Camera move ignored	Too busy for one shot	Split it into two cameras, one move each
Faces change between shots	No shared reference	The same reference sheet and description on every shot
Mannequin look leaks in	The prompt leaves the look open	Describe the people, clothing, materials and light
Characters seem to swap sides	The camera crossed the line	Check the timeline warning and move the camera back

Perguntas frequentes

Can I use a render from another 3D tool?

Yes. Upload a still or an MP4 rendered in another 3D tool and attach it as the reference. Blender users can skip the export: the free Morphic for Blender add-on sends the last render or animation to Morphic in one click. You can also upload .glb, .gltf or .obj models into 3D Scene.

Which video models work best with a 3D reference?

Pick a video model that accepts a reference video when the camera move matters; the model picker shows what each model accepts. When one strong still is enough, generate a finished frame from the blocking snapshot and animate it with image to video. Run two models side by side and keep the take that follows the blocking best.

Will the generated video follow my camera move?

Yes, on simple moves. The blocking video guides the camera, the staging and the screen direction, and models follow it closely. Fast or complex moves can drift, so keep each shot to one clear move and generate a couple of takes. Treat the blocking as direction the model reads, then keep the take that matches it best.

Can I keep the same character across shots?

Yes. Make a reference sheet for each character and attach it to every shot, alongside the blocking video, then name the figure by colour in the prompt. The blocking places the person and the reference sheet carries the face, costume and look, so the same character holds from the wide to the close-up.

Do I need 3D experience to block a scene?

No. Describe the shot to Copilot and it builds the set, cast, action and camera in 3D Scene for you, then you adjust anything by hand. The editor runs in a desktop browser, renders the blocking video at no credit cost, and keeps every object editable, so you can learn it by changing what Copilot made.