Lightricks, primarily known for its image and video editing software, just announced a preview of its open-source AI video model, LTX Video. This model makes possible medium-resolution, real-time video generation, a feature available until now only in a handful of large language models (LLMs) but now in an open-source release. Once freely released, LTX Video will be free of charge for personal and commercial usage and can be integrated into the LTX Studio platform.

LTX Video produces five-second videos, which come in a resolution of 768 x 512 pixels. This model accepts both text and image inputs. The preview version restricts video quality to medium resolution but allows for rapid generation times, with a four-second wait time. This performance is enabled on devices carrying the Nvidia H100 chip. The model was designed on the basis of the Diffusion Transformer architecture but is streamlined to use only two billion parameters so that it can keep its file size small.

Lightricks said the AI model generates dynamic videos that have a high adherence to prompts and could be run on systems with GPUs like an RTX 4090. At the moment, it is available for download via GitHub, Hugging Face, and ComfyUI. It also has a wide compatibility with external editing tools, through which users can fine-tune the resulting video. The company will launch LTX Video, the full version of the product, which will be open-source and available for personal as well as commercial use. What's more, it will be launched with LTX Studio, an AI-powered storyboard tool.

Read More News :

Mphasis Partners with Haas F1 to Revolutionize Sports Tech

Ascendion Unveils New AI Studio in Hyderabad to Boost Innovation