Robotics & Physical AI
Open world models for physical AI
Train on worlds you generate. Deploy on hardware you choose. Our open world model is built for synthetic data, simulation, and physical AI, fine-tuned on your environments and running on your infrastructure.
.svg)
Downloads on HuggingFace
Runs locally on 12GB VRAM
Near real-time generation
Why LTX
**From simulation** to action
One open foundation for every stage of the physical AI journey, from simulation to world-action models.
Generate synthetic data
Create diverse environments, viewpoints, lighting conditions, and object interactions to expand robotics training datasets.
Adapt to your environment
Fine-tune LTX on proprietary environments, sensor data, and robotics tasks using IC-LoRA.
Deploy with efficiency
Run inference in near real time on a single GPU: on-prem, in your cloud, or at the edge where robots actually operate.
Build toward world-action models
Start with synthetic data and simulation today. The same open foundation is built to extend toward real-time policy generation, without switching stacks when you get there.
Teaching robots how the world behaves
"Training robots means teaching them how the physical world actually behaves, not just what it looks like. LTX-2.5 is the open model that gets closest to that for us, and being able to run and fine-tune it on our own hardware is what makes it usable for real robotics work. We have not found another open model that does this for us the way this one does."
Capabilities
**Built for** real-world physical AI
Physical understanding
Generate motion, interactions, and real-world dynamics to produce simulation-ready robotics data.
Controllable generation
Guide generation with motion, depth, pose, camera, and other conditioning signals, so every trajectory matches what you need to train on.
Pretrained for deep adaptation
Start from a pretrained foundation designed for extensive fine-tuning. Adapt LTX to your environments, sensors, and robotics tasks without rebuilding the model from scratch.
Security & IP
**Built for** enterprise ownership
Deploy LTX with the flexibility enterprises need, from permissive licensing and open deployment to full ownership of your IP and infrastructure.
Deploy on-prem or in your private cloud.
Fine-tune, deploy, and commercialize under permissive licensing.
Run on any hardware, with no vendor lock-in.
Keep models and IP under your control with self-hosted deployment.
Research
Explore the research behind LTX
Go beyond the demos. Discover the research, architecture, and technical breakthroughs behind LTX's open world models.
Explore
Use cases
See the features that turn the model into a production pipeline. Built for scale, control, and integration.
Resources
**Learn more about** LTX for robotics
Explore how LTX helps robotics teams generate synthetic data, fine-tune world models, and build the foundation for physical AI.
Introducing LTX as an Open World Models Company
.jpeg)
Why Train Your Own AI Video Model? Control, IP Protection, and Custom Pipelines
Open Source Video Generation Models (2026 Landscape Guide)
Training Your First LoRA on LTX
.jpeg)
Using LoRA Adapters with LTX-2.5: A Developer Guide
Hardware Guide for AI Video Models: GPU Recommendations by Use Case
