You can now call Uthana's text-to-motion and auto-rigging models with Replicate through the API or instantly in the browser playground. No Uthana account required. A few lines of Python or Javascript and you're generating animation or rigging characters programmatically.
We're launching three models:
All three output standard FBX and GLB files that work in Unity, Unreal, Blender, Maya, and any pipeline that imports skeletal animation.
Replicate lets you run AI models through a simple API with managed infrastructure — no GPUs to provision, no containers to manage, pay only for the compute you use. If you're already building with Replicate for other models, adding Uthana's animation and rigging models is the same replicate.run() call you already know. If you're new to Replicate, it's one of the fastest ways to go from "I need animation generation in my project" to working code.
Uthana's own platform at uthana.com offers capabilities that go beyond what's available on Replicate — retargeting to your own character rigs, a 100,000+ searchable motion library, DCC plugins, video-to-motion, and tools for stitching, blending, and looping clips. If your workflow needs those, the full platform is the right path, working either within our webapp or API. But for calling the core models programmatically, Replicate gives you a fast, familiar way to do it.
Our autoregressive motion generation model. It takes a text prompt and produces a skeletal animation clip on a standard bipedal rig. 1.0 is designed for speed and simplicity. Two parameters: your prompt and an optional foot IK toggle for cleaner ground contact.
Our diffusion-based motion generation model. Same input, a text prompt, but a different architecture underneath. Where 1.0 predicts motion token-by-token, 2.0 starts from noise and iteratively refines it into a complete motion sequence. It also exposes more control: diffusion steps (quality vs. speed tradeoff), classifier-free guidance scale (prompt fidelity vs. diversity), output length in frames, and a seed for reproducibility. The seed parameter is especially useful in automated pipelines, allowing you to lock in a result you like and reproduce it reliably across runs.
The two models are complementary.
Upload a bipedal humanoid mesh in FBX or GLB (T-pose or A-pose, feet on the ground) and get it back with a complete skeletal rig. The model analyzes the mesh geometry and places bones to match the character's proportions — hips, spine, shoulders, elbows, knees, all of it — in under 30 seconds.
This is useful if you're building a character pipeline that takes meshes from 3D generation tools, marketplace assets, or user uploads and needs to rig them automatically before animation. The output is a skeleton (no skinning or weight painting), which is enough to apply animation via retargeting. For production characters, treat it as a fast starting point and refine the rig in your DCC.
Each model's page on Replicate includes an API tab with ready-to-use code in Python, JavaScript, and HTTP. If you're already using Replicate for other models, the pattern is identical — install the SDK, pass your inputs, get your outputs.
A few things worth knowing for integration:
auto-rig-1.0, then animate it with either text-to-motion model. The output of one feeds directly into the next.text-to-motion-2.0 accepts a seed parameter. Set it to get deterministic output — useful for testing, CI pipelines, or any workflow where you need consistent results.text-to-motion-1.0 in a loop with different prompts to generate an animation library programmatically. The model's fast inference and simple parameter set make it well suited for high-volume use.See each model's API tab for the full parameter reference and code examples:
Replicate gives you access to some core models — text in, animation out; mesh in, rigged character out. The full Uthana platform adds everything around those models that a production pipeline needs:
If Replicate is the fast path to calling our models, uthana.com is the full platform for building with them. You can sign up for a free account and get 100 seconds of downloadable animation per month to start.
If you build something with these models, we'd love to see it — reach out on Discord or drop us a line at support@uthana.com.
Start generating motion today, or work with us on a custom dataset, embodiment system, or simulation program.
