LTX.io

Reformat one master video for every platform and ratio, and batch a full campaign without opening an editor.

Outpaint 16:9 into 9:16, 1:1, or 4:5 without cropping your subject. Send your next batch through the API.

Native workflows, multishot continuity, and generative editing, all built for finishing-grade output. Learn how Comfy are building with LTX.

Zooming in cuts people out of frame. Outpainting keeps your footage whole in any ratio, ready for your next placement.

Had a great conversation with Deirdre Bosa and Lukas M. Ziegler on Nebius Live at NVIDIA GTC Berlin about why video models are becoming the foundation for robotics. The short version: LLMs scaled abstract intelligence. They prove theorems, they write code. But they have no intrinsic sense of time or space. Video models do. Train one on enough of the real world and it becomes an intuitive physicist. It won’t hand you equations, but it builds a convincing simulation of how things move and interact. That changes robotics. Instead of programming every task, a robot looks at a scene, gets an instruction, imagines the next half second, and acts. Because the model has seen humans do these things thousands of times, it can zero-shot tasks it was never trained on. We also talked about where robots land first. My bet: construction sites and factories before kitchens. Not because homes are less valuable, but because safety is far easier to solve where there are no kids in the room. And on form factor: humanoids will matter, but so will dogs, drones and arms. The interesting work is the layer that transfers across all of them.

Personalize it into new styles, characters, and capabilities, without losing control of the model itself. Learn how Fal are building with LTX.