UNLIMITED Seedream 5.5 · Live on every paid plan

See how
A glossy pink product bottle on a white pedestal in a bright 3D studio set
3D planning

Official BytePlus partnerBytePlus, a ByteDance company

AI previs: plan shots in 3D before you make the video

What AI previs is and how to do it: stage characters and camera in 3D, fix the shot by chat, then hand a clean start frame to an AI video model. Free to plan.

KreatorFlow4 min read

Short answer

AI previs means staging a shot in 3D before you make the video: where the people stand, where the camera is and how it moves. It fixes the composition so an AI video model is not guessing it from text. In Stage you describe the scene to Kiryu, fix it by chat, write the camera move from 40 named moves, and export the previs video and start frames. All of that costs no credits. Only turning the shot into an AI video needs a paid plan.

Updated:

Do AI previs in six steps

  1. 1

    Write the shot in one sentence

    Who is in it, where, what happens and how the camera moves. One shot at a time.

  2. 2

    Open Stage and describe it to Kiryu

    Go to Stage. You can start without an account. Pick one of 22 examples or type your idea. Kiryu builds the scene in 3D in your browser.

  3. 3

    Fix it by chat

    Say make it night, move her left, add a crowd or make the shot wider. Building and every edit you type spend no credits, within daily fair-use limits.

  4. 4

    Write the camera move

    Pick a framing (11 of them), a lens (14, 24, 35, 50, 85, 135 mm) and a camera move such as push in, orbit, crash zoom or vertigo. Each move writes keys you can drag on the timeline.

  5. 5

    Play it and check the timing

    Press Space. Each shot can be up to 15 seconds and a sequence holds up to 36 shots.

  6. 6

    Export, then decide on AI video

    Export the previs video and clean start and end frame PNGs, all free. If the shot works, Make it real hands it to Seedance 2.5. Making the video needs a paid plan; free accounts can plan, preview and export the previs.

What previs is

Previs, short for previsualisation, is a rough 3D version of a shot made before the final footage. Film crews use it to decide camera positions and timing. For AI video it does a similar job: the model sees pictures and text, not your 3D scene, so a framed start frame and a clear brief carry your plan to it.

Why plan before you generate

AI video models make the motion themselves, so a vague prompt gives a different shot each time. A planned start frame fixes who stands where and how the shot is framed. You still treat the camera move as strong guidance, not a locked path.

Tools for AI previs

You can block a shot in a full 3D suite such as Blender, with or without an AI assistant driving it through MCP. That gives you full control and a long setup. Stage is the shorter route for this one job: it runs in the browser, builds from menus and keeps the camera moves editable.

Limits of Stage previs

Scenes are built from menus of sets, props, characters, creatures and moves. There is no free modelling, sculpting, UV work or node materials, and it renders in real time, not path-traced. You can import your own .glb, .gltf or .vrm model. Free and signed-out exports carry a small "Staged with KreatorFlow" mark.

In Simple mode, Make it real sends Stage's own render of every shot to Seedance 2.5 (or to Kitsune, which keeps Stage's exact motion) and makes the whole video at 720p, showing the exact credits before you confirm.

Frequently asked questions

What is AI previs?

AI previs means staging a shot in 3D before you make the video: where the people stand, where the camera is and how it moves. It fixes the composition so an AI video model is not guessing it from text. In Stage you describe the scene to Kiryu, fix it by chat, write the camera move from 40 named moves, and export the previs video and start frames. All of that costs no credits. Only turning the shot into an AI video needs a paid plan.

Is previs free in Stage?

Yes. Opening Stage, building scenes with Kiryu, editing, camera keys and exporting the previs video cost no credits. Turning a shot into an AI video needs a paid plan and spends credits after the cost is shown.

Do I need Blender for previs?

No. Blender is one way to do it. Any tool that fixes the composition and the camera works, and Stage does it in the browser.

Can I use my own character?

Yes. Import a .glb, .gltf or .vrm model up to 50 MB. It stays in your browser.

Does the AI video follow my camera move exactly?

It uses your framed start frame and your camera and action in an editable brief. The model makes the motion itself, so treat the move as strong guidance.

Plan your next shot in 3D

Open Stage, describe the scene to Kiryu and fix it by chat. Planning is free.

More guides