LTX-2.5 is a new-generation open-weight audiovisual generation model from LTX, capable of simultaneously generating high-quality video and synchronized audio in a single pass. It supports text-to-video, image-to-video, and start/end-frame-guided video generation, while delivering enhanced visual detail, motion stability, comprehension of complex prompts, and consistency across long-sequence content.Compared to LTX-2.3, LTX-2.5 introduces an all-new Diffusion Video Decoder that effectively mitigates issues such as blur, texture degradation, and loss of character detail during high-speed motion. It also incorporates native multi-shot generation capabilities, enabling the creation of multiple consecutive shots in a single run while maintaining—to the greatest extent possible—consistency in characters, scenes, lighting, style, and audio across shot transitions.The model utilizes an LTX-customized Gemma 4 12B text encoder, offering superior comprehension of multiple characters, actions, cinematic language, lighting, and complex scene descriptions. It also supports automatic duration prediction, dynamically matching the video length to the actions and scene content specified in the prompt. Furthermore, LTX-2.5 supports native 4K HDR, frame rates up to 50 FPS, and RAW workflows, providing greater flexibility for post-production in film, television, and high-quality video creation.As an open-weight model, LTX-2.5 supports local deployment, LoRA, and fine-tuning. Balancing generation speed, controllability, and openness, it is well-suited for applications such as short films, advertisements, character animation, multi-shot storytelling, and professional video production.
