LTX, the open world model company, has released LTX-2.5, the newest version of its open-weights video and world model. The model is available now as open weights on Hugging Face, inside ComfyUI, and through the LTX API for teams that want managed generation. It is free to use for organizations under $10 million in annual recurring revenue; larger companies negotiate a license.
LTX-2.5 Capabilities
LTX-2.5 rebuilds nearly every stage of the generation pipeline, with a new diffusion video decoder, native multishot generation, and a custom Gemma 4 language backbone. The model also features a pretrained checkpoint tuned for physical AI and robotics, giving teams a base to fine-tune on domain data.
They claim LTX-2.5 generates a 10-second, 720p image-to-video clip in 6.8 seconds faster than real time, measured self-hosted on two of NVIDIA’s top-end GB200 chips. The model’s speed and quality make it an attractive option for enterprises and developers.
Open Weights and Licensing
LTX-2.5 is available for use, and it is free to use for organizations under $10 million in annual recurring revenue; larger companies negotiate a license.
Partnership with ComfyUI
LTX has partnered with ComfyUI, a node-based workflow tool, to provide zero-day support for the Comfy integration. ComfyUI is a popular prototyping system in the industry, and many of LTX’s paying customers start their journey with Comfy.
Enterprise Adoption
Both LTX and ComfyUI executives pushed back on the assumption that a video model is only for generating videos. LTX’s CEO, Zeev Farbman, listed enterprise deployments that have little to do with cinematic clips, such as computational photography, VFX, and animation studios.
Related: OpenAI unveils GPT‑5.6 Cyber with 95% success
For enterprises weighing where to start, the recommended path is to adopt ComfyUI, which provides a structured environment for tweaking and abstracting away complex tasks. Enterprises can then reach out to LTX for licensing and support.
Real-Time Applications
LTX-2.5 is designed to be efficient and run on edge devices, both on consumer hardware and close to the edge with physical AI. The model’s performance and low latency make it suitable for real-time applications, such as interactive avatars, live worlds, and robotics workloads.
Virtual Production and Filmmaking
For professional filmmakers and studios, real-time world models are changing the shape of production itself, collapsing the gap between shooting and post. Virtual production allows directors and producers to see close to the final result immediately, streamlining the pipeline and reducing iterations.
LTX-2.5 is available today on Hugging Face, natively in ComfyUI, and through the LTX API.
The field of video and world models continues to evolve.
LTX is betting on a strategy that prioritizes openness, efficiency, and flexibility, and it will be interesting to see how this approach plays out in the market, as companies like LTX and ComfyUI pass security checks and gain traction.
