Guides
Can ChatGPT animate a 3D model? What works, and what's a video
Can ChatGPT animate a 3D model? Only through tools: Vi3W rigging and moves over MCP, or Blender keyframes in Python. Steps and limits.
5 min read

Quick answer
Can ChatGPT animate a 3D model?
Not on its own. ChatGPT writes text, code and images, so a 3D animation needs a tool it can drive. Connected to Vi3W over MCP, it can rig a model and apply moves like walk, run, jump or wave. Through Blender, it can write Python that sets keyframes. An animated photo is a video, not a 3D animation.
Two things people mean by "animate"
Search "ChatGPT animate" and most results are about animating photos: a portrait that blinks, a pet that wags. That's video. It's a sequence of flat frames, and you can't drop it into Unity.
Animating a 3D model is different. The model needs a rig (bones and skin weights), and the animation is data on those bones: rotations over time, saved in a GLB or FBX. A game engine plays that data on the mesh from any angle.
ChatGPT can help with both, but neither happens inside ChatGPT. It drives tools. This guide covers the 3D kind. If you want to turn a picture into a moving 3D character, see animating a picture in 3D.
Route 1: rig and animate through the Vi3W MCP server
With Vi3W added to ChatGPT as a custom MCP server, the steps are short requests:
- Generate or pick a model. "Make a stylised fox character, standing on all fours, plain background." Or use one you already have in Vi3W.
- Rig it. "Add bones to the fox. It's an animal." The animal option gives a creature rig. Humanoid is the default.
- Animate it. "Make it walk." Make it move offers Idle, Walk, Run, Jump, Wave and Dance, or you can describe a motion.
- Download. "Give me the animated fox as FBX for Unity."
When you describe a motion ("a happy little hop", "throw a punch"), Vi3W matches your words to its library of motion clips. It's not inventing new motion from scratch, so a description that matches a known move works, and something unusual ("a sad slow-motion moonwalk") may get "we couldn't find that move, try words like walk, run, jump, wave, dance or kick." Rephrase in simpler terms.
Two practical notes: Walk and Run come with every rig, so they cost only the rig. And each move you add rigs the model again, so finish shape and texture changes first.
On plans: full MCP with write actions in ChatGPT is currently for Business and Enterprise/Edu workspaces on the web, and Pro gets read and fetch only, per OpenAI's article. Claude's custom connectors work on Free and paid plans. The steps there are in can Claude rig a character.
Route 2: keyframes in Blender, written by ChatGPT
Without a connector, ChatGPT can still write Blender Python that animates. You paste the script into Blender's Scripting workspace and run it. For simple object animation this works well:
- a coin that spins and bobs,
- a door that swings open over 24 frames,
- a camera orbit around a product,
- a bouncing ball with squash and stretch.
A good prompt names the object, the frame range, the frame rate and the curve:
"Write a Blender Python script that animates the object named 'Coin': rotate 360 degrees on Z from frame 1 to 60, move up 0.2 m and back down over the same range, set the scene to 30 fps, and make the rotation interpolation linear so it loops cleanly."
In Blender, I inserts a keyframe on whatever you've selected. The script does the same thing in code with keyframe_insert. If you're connected over Blender MCP (in Claude, Codex, Cursor and others), the assistant can run that code directly instead of making you paste it.

Characters through code: possible, rarely good
ChatGPT can keyframe bones on a rigged character, so "raise the right arm, then wave it three times" works as code. A believable walk cycle is another matter. It's dozens of coordinated rotations with weight shift and overlap, and code-written walks look robotic. For character motion, use captured or library motion (Route 1, or Mixamo for humanoids) and use code for the simple stuff. AI walk cycles from a prompt goes deeper.
Route 3: Mixamo, with ChatGPT doing the prep
For humanoids, Mixamo's free library is still a strong option. ChatGPT can write the Blender script that prepares and exports your character (joined mesh, applied transforms, FBX with embedded textures). You upload and pick animations on Mixamo, then bring the FBX back. Mixamo only rigs bipedal humanoids, so creatures need Route 1 or a hand rig.
Checking an animation is actually usable
Whatever made it, check these before you build on it:
- Feet sliding. In a walk, planted feet shouldn't skate. In-place walks are normal for games, where the character controller moves the root.
- Mesh tearing at joints. Shoulders, hips and knees are the usual places. That's a weight problem, not an animation problem. See fixing weight painting with Claude.
- Loop seams. Play the clip on repeat. A hitch at the loop point means the first and last frames don't match.
- Frame rate and length. Engines expect specific rates. Tell the assistant your target (30 fps is common for game clips).
- Format. FBX for Unity and Unreal, GLB for web and Godot. Check which 3D formats carry animation.
What ChatGPT can't do with 3D animation
- Act out a performance. Lip-sync, facial expressions and a hand-keyed emotional beat are an animator's job. No route here gives you that.
- Invent new motion from a paragraph. Text-to-motion research exists, but the practical tools today pick and adapt motion from libraries.
- See the animation. It can't watch the playback. You have to.
- Fix a bad mesh by animating it. A character with fused arms will animate with fused arms. Generate it in a clean A-pose or T-pose first.
Try it on one character
Generate a character in a neutral pose, connect Vi3W to ChatGPT or Claude, and ask for Add bones then Make it move with a walk. Open the result in Blender and press play. Open Vi3W.
Frequently asked questions
can claude animate 3d models
The same way: through tools. Claude with the Vi3W MCP server can add bones and preset moves. Claude with Blender MCP can keyframe objects and bones in Python. Neither model animates by itself.
does chatgpt animate
ChatGPT can make images and, with video tools, short clips from images. Those are flat video. For a 3D model that moves in a game engine, you need a rig and animation data, which come from connected 3D tools.
ai animation 3d model
AI can animate a 3D model by auto-rigging it and applying motion from a library of captured moves. The output is a rigged GLB or FBX with animation clips that play in Blender, Unity, Unreal or Roblox.
Last reviewed

