LTX Video 2.0 19B

• 🆕 Lightricks' open-weights image-to-video model, released January 2026.
• 🔧 Full-quality variant of the 19B DiT architecture, optimized for fidelity.
• 👁️ Animates a source still image into synchronized video plus audio.
• 🎯 Generates synchronized video and audio in a single pass.
• 📏 Targets up to 4K resolution, 50 FPS, and 20-second clips.
• 🔒 Released under the permissive Apache 2.0 license with full weights.
• 🌐 Runs locally via ComfyUI or NVIDIA-optimized deployments.

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 2.0 19B is the full-quality, image-to-video member of Lightricks' LTX-2 family, an open-source generative video system released in January 2026. It uses a 19-billion-parameter diffusion transformer (DiT) architecture and, according to Lightricks, generates synchronized 4K video and audio in a single pass, supporting clips up to 20 seconds at 50 FPS. This full-quality configuration prioritizes fidelity over speed, rendering more slowly than the distilled or fast checkpoints.

Compared with the earlier LTX Video 2.0 Full Quality release, the 19B variant scales up the parameter count and folds joint audiovisual generation into one workflow rather than requiring separate audio post-production. Lightricks distributes complete model weights, training code, and the inference pipeline under the Apache 2.0 license, alongside LoRA fine-tuning support for character and style consistency.

A faster sibling, the distilled LTX Video 2.0 19B Distilled, trades some quality for rapid iteration. Lightricks later shipped LTX Video 2.3 Full Quality in March 2026 as a subsequent update within the same family.

The full model weights, configuration, and inference code are published on Lightricks' Hugging Face repository, enabling local deployment through tools such as ComfyUI. Within the catalog, this image-to-video checkpoint is paired with a matching text-to-video build, giving creators both prompt-driven and image-conditioned generation from the same 19B base.