LTX Releases LTX-2.5 Video Generation AI as Open Model
LTX-2.5 extends the company's open-model approach with multi-shot and HDR capabilities, positioning it against proprietary rivals while keeping local and API-based generation accessible.
Reporting from 1 source: GIGAZINE.
LTX released LTX-2.5, an open video generation AI model, on August 12, 2026. It supports text, image, and video inputs, generates videos with audio, and natively handles multi-shot generation. The model uses a Gemma 4-based custom LLM for prompt adherence and includes Diffusion fidelity rendering for quality. Commercial use is allowed for companies with annual revenue under $10 million.
LTX has released LTX-2.5, its latest video generation AI, as an open model. The model generates videos with audio from text, image, or video inputs, and natively supports multi-shot generation, preserving face and voice continuity across multiple shots in a single pass.
The company says the model follows complex instructions for motion, lighting, text, and composition, thanks to a custom LLM based on Gemma 4. A technique called Diffusion fidelity rendering allocates compute based on scene complexity, and the model supports 16-bit HDR output.
LTX-2.5 is available on Hugging Face, and ComfyUI supports it. The company says it can run on a GPU with 16GB of VRAM, and LoRAs trained for LTX-2.3 remain usable.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.