LTX-2.5 🧠 Model Open source

What it is

LTX is Lightricks' open-weight video family. The company also makes LTX Studio and Facetune. LTX-2 (January 2026) was the first open model to generate synchronised audio and video in one pass. LTX-2.3 (March 2026) came with a local desktop editor, and LTX-2.5 (August 2026) adds native multi-shot generation and a new video decoder. Weights, inference code and training code are public.

Key features

• Synchronised audio and video in one pass
• Native 4K (HDR on 2.5) at up to 50 fps. About 5 seconds per clip at default settings (121 frames)
• Multi-shot generation that keeps characters consistent across cuts
• Runs locally on consumer GPUs (16 GB VRAM minimum for 2.5)
• Text-to-video, image-to-video, audio-to-video and LoRA training

Where you can use it

Download from Hugging Face (Lightricks/LTX-2.5) and run it in ComfyUI (day-one support and LTX-2.5 partner nodes). You can also use Lightricks' LTX API or LTX Studio. Krea offers LTX-2 and LTX-2.3.

Pricing and rights

The open weights are free for academic use and for commercial use by organisations under $10M in annual revenue, under the LTX-2.x Community License. Bigger companies need a paid licence. LTX API prices per generated second: $0.09 at 720p, $0.15 at 1080p, $0.19 at 2K and $0.37 at 4K.

Who it is for

Developers, ComfyUI users and small studios that want a local, controllable model with sound, and no per-clip fees.

Verdict

LTX-2.5 is the most complete open audio-video model and fast enough to use for real work. The licence is not open source in the OSI sense. Its looks are still a step behind the closed leaders like Veo and Kling.