LTX-2 / Veo3 open-source video model

Lightricks open-sourced LTX-2, an LTX2 video model comparable to Veo3. - LTX-2 generates up to 20s synchronized video with audio and lip-sync in high resolution. - The model accepts text, images or rough sketches as prompts and runs fast enough for practical use on common consumer GPUs. - Authors claim generation speed and quality exceed some closed-source alternatives, advancing open-source AI video capabilities.
Key insights
LTX-2 Delivers Fast, High-Quality 20-Second Multimodal Video Generation
Posts report that LTX-2, released by Lightricks, can produce up to 20-second videos with synchronized audio and lip motion from text, images or sketches. The model emphasizes speed and quality improvements over prior open and closed alternatives and is optimized to run on commodity GPUs, enabling broader experimentation and potential integration into content workflows.
Open-Source Video Models Expand Practical AI Video Production
The open-source release of LTX-2 is positioned as accelerating adoption of AI-generated video by lowering entry barriers for developers and creators, providing multi-modal conditioning and improved throughput versus earlier models.