Data Augmentation

Expand your training distribution without collecting more data. GWM-1 generates photorealistic variations of your existing trajectories — new environments, lighting, camera angles, and object placements — so you can scale training data without scaling data collection.

Key benefits

More data, without more collection.

Scale without collecting

Turn a handful of demonstrations into hundreds of training examples. GWM-1 generates realistic variations that preserve the task structure while diversifying the visual distribution.

Video-to-video generation

Transform existing trajectory videos into new environments and conditions. Change backgrounds, lighting, add obstacles — while maintaining the same action sequence.

Image-to-video generation

Generate full trajectory videos from a single starting frame and a text prompt. Explore how your robot would behave in entirely new scenarios.

Sign up to access the Data Augmentation SDK and start expanding your training distribution with GWM-1.

Get Access