This company develops an Open Creativity Stack, an ecosystem of AI tools and models designed to empower creators, studios, and enterprises. They build and deploy efficient AI models, integrating them into proprietary products for visual storytelling and content creation. Their offerings range from mobile editing apps like Facetune and Videoleap to advanced AI video generation platforms.
Funding
Funding not disclosed
Founders
Product
Problem
Generating long-form video content with AI often results in inconsistencies, flickering, and high processing times, hindering efficient and scalable video production. Existing solutions struggle to maintain visual quality and coherent motion across extended video sequences.
Solution
Lightricks offers LTXV, an open-source AI video generation model designed for scalable long-form video production with high visual fidelity and consistent motion. LTXV leverages a transformer-based latent diffusion architecture and multiscale rendering techniques to ensure coherent transitions between frames, addressing common issues like flickering and inconsistencies. Optimized for both GPU and TPU systems, LTXV significantly reduces video generation times while maintaining high visual quality. The model supports multiple input types, including text-to-video and image-to-video, providing flexibility for diverse video creation workflows. LTXV is available as a standalone tool or through LTX Studio, Lightricks' all-in-one platform for video production.
Target Audience
The primary audience includes video content creators, researchers, and developers seeking an efficient and customizable AI video generation solution.
Features
- Open-source AI video generation model built on a transformer-based latent diffusion architecture
- Multiscale rendering technology for sharp, smooth videos with fewer artifacts
- Frame-to-frame learning to ensure coherent transitions and eliminate inconsistencies
- Optimized for consumer-grade hardware, including NVIDIA RTX 4090 and 5090 GPUs with as little as 8GB of VRAM
- Supports multiple input types: text-to-video, image-to-video, keyframe animation, and video extension
- LoRA-based training and multi-GPU support for custom fine-tuning
- Available in distilled and quantized versions for faster performance on limited hardware
- Integration with ComfyUI and Hugging Face Space for workflow examples and testing