LTX Video 2.5 Fast

• 🎬 Text-to-video model from Lightricks, fast tier of LTX-2.5
• 🧠 Built on a 22-billion-parameter asymmetric dual-stream diffusion transformer
• ⚡ Distilled checkpoint runs in a handful of denoising steps
• 🆕 Succeeds LTX Video 2.3 Fast in the same speed-oriented tier
• 🔧 Family adds retake, keyframe interpolation, and IC-LoRA control adapters
• 🔒 Open weights under a permissive licence with commercial terms
• 🎯 Aimed at iteration, storyboards, and preview renders
• 📚 Training code, ComfyUI nodes, and reference workflows published

Lightricks

Lightricks is a creative technology company founded in January 2013, widely recognized for its popular selfie-editing app Facetune. Originally focused on mobile video and image editing tools, the company has expanded significantly into generative AI,…

LTX Video 2.5 Fast is the speed-optimized text-to-video tier of Lightricks' LTX-2.5 generation, released in 2026. Lightricks positions the LTX family as an open, production-grade system for synchronized audio–video creation, with LTX-2.5 built on a 22-billion-parameter asymmetric dual-stream diffusion transformer. The Fast tier is the distilled checkpoint, which Lightricks describes as running in a handful of steps where the full model takes twenty or more — the reason it is recommended for rapid iteration, storyboards, and previews rather than final renders.

Compared with its direct predecessor, LTX Video 2.3 Fast, Lightricks now lists 2.3 as the previous generation and 2.5 as the recommended starting point for new projects, specifically citing multi-shot continuity, editing existing footage, and domain adaptation as areas where 2.5 is the better fit. The 2.5 line also broadens the surrounding toolkit: audio-to-video and text-to-audio modes, keyframe interpolation, retake for regenerating a single stretch of a clip, dialogue replacement, and IC-LoRA adapters for pose, depth, inpainting, relighting and upscaling. One practical migration note from Lightricks: custom LoRAs must be retrained for the new latent space.

Within the catalog it sits alongside LTX Video 2.5 Fast for image conditioning and the higher-fidelity LTX Video 2.5 Pro, letting teams prototype on the Fast tier and finish on Pro through the same interface. Weights, inference code and ComfyUI workflows are published openly.