LTX-2.5: Lightricks' open video generation model — 10-second clip in 6.8 seconds

🛠 LTX-2.5: Lightricks' open video generation model — 10-second clip in 6.8 seconds

The new diffusion video decoder removes artifacts, native multishot generation creates coherent frames while preserving the character, the language core is a custom Gemma 4, and it integrates with ComfyUI.

🌍 Generating 10 seconds of 720p in 6.8 seconds — faster than Gemini Omni Flash (52 seconds), Grok 1.5 (63 seconds), and Veo 3.1 (70 seconds). Blind tests: 67% win rate. The LTX family has 33 million downloads.

👤 Locally on a 16 GB GPU, free for organizations with <$10M ARR, integrates with ComfyUI, no watermarks.

Source 1: https://ltx.io/model/ltx-2-5 Source 2: https://cryptobriefing.com/ltx-2-5-ai-video-model-release/