LTX-2.5
#8 in Text-to-VideoLightricks · v2.5 · 7× · last seen Aug 13, 2026
67
Momentum
LTX-2.5 is an Open-Weights audio-video foundation model from Lightricks (LTX) based on a 22-billion-parameter Diffusion Transformer that generates video and audio together in a single pass. It supports native multi-shot generation with consistent characters, scenes, and voices across multiple shots, as well as 4K-HDR output and automatic clip length determination. The model can be run locally on your own hardware, Fine-Tuned, and accessed via ComfyUI, Hugging Face, or the LTX API.
Momentum trend
15.05.13.08.
Features
| Fine-Tuning | Yes, dedicated pretrained base checkpoint for fine-tuning and deployment on your own infrastructure |
| Generation Time | 10-second clip in 6.8 seconds (self-hosted on 2x GB200 GPUs, 720p) |
| License | LTX-2.x Community License (free for companies <$10M revenue; commercial license required above that) |
| Max Resolution | 4K (Fast variant); Pro variant tops out at 1080p |
| Max Video Length | Up to 20 seconds (Fast variant, 2-20s); Pro variant 2-10s |
| Platform | ComfyUI, Hugging Face, GitHub, LTX API, LTX Desktop, Runway |
| Price | Free for organizations under $10M annual revenue (Community License); API from $0.09/s (720p) up to $0.37/s (4K) |
| Release Date | August 11, 2026 |