Video Gen · Model Guide · 4 min read
LTX-Video 2B
The speed demon of open video — a 2B model that generates clips faster than they play back, on 8–12 GB cards.

| Developer | Lightricks |
|---|---|
| Availability | Open weights |
| Commercial license | Open weights under Lightricks' release license (check the current model card — terms have shifted between versions, with some releases non-commercial). The 2B base model's release was broadly permissive; verify the exact repository you download. |
Basic Information
LTX-Video is Lightricks' compact open video model built around one idea: real-time-scale generation. Its small 2-billion-parameter DiT design produces short clips dramatically faster than its larger siblings — near-playback speed on capable GPUs — at the cost of lower peak fidelity than Wan or Hunyuan.
For interactive workflows (rapid iteration, storyboarding, live previews) and modest hardware, it is the open video model to reach for, and newer releases continue improving resolution and duration.
Best use cases
- Fast iteration: many prompt variations in the time one big model generates one clip
- Local video generation on 8–12 GB consumer GPUs
- Real-time-ish previewing inside ComfyUI pipelines
- Learning video-diffusion workflows on affordable hardware
Computational Requirements
The most accessible video model on this list: roughly 5–10 GB of VRAM depending on resolution and quantization — entry-level cards need not apply elsewhere.
| Configuration | Memory |
|---|---|
| FP16 (reference) | ~4 GB |
| Practical target | 8–12 GB cards for 720p-class output |
Commercial license
Open weights under Lightricks' release license (check the current model card — terms have shifted between versions, with some releases non-commercial). The 2B base model's release was broadly permissive; verify the exact repository you download.
Check whether it fits your hardware
LTX-Video 2B is in our compatibility library — open the compatibility panel and look it up with your GPU and context length.
More guides: back to Video Gen · visit the blog
