A LongCat Video generator turns a written prompt or a single still image into a moving clip — and, unlike most hosted tools, it keeps going. The generator on this site runs the open-weights LongCat Video model on managed GPUs, so you get minutes-long, coherent output in the browser without a local GPU, a ComfyUI graph, or a folder of model weights.
Three things, in one workflow:
Output targets 720p at 30fps, which is a practical format for YouTube, explainers, social edits, and concept footage — not just a preview loop.
No seed archaeology, no per-frame colour matching, no timeline surgery.
| Hosted LongCat Video generator | Local install (ComfyUI / Pinokio) | |
|---|---|---|
| Setup | None — sign up and generate | Python, CUDA, checkpoints, workflow wiring |
| Hardware | Managed GPUs | Your own VRAM budget |
| Best for | Getting a usable clip now | Full control, offline runs, custom nodes |
| Long output | Continuation built in | Manual continuation and stitching |
If you want the model weights, commands, and VRAM numbers, the local setup guide is the place to start. If you want to see what the model can do before installing anything, generate here first.
Both spellings find the same model — Meituan's open-weights LongCat Video, a 13.6B-parameter video model. The space-separated spelling is just how many people type it; the long cat video page covers that spelling directly. The model itself is unrelated to LongCat Flash, which is a language model, not a video model.
Yes. Create a free account and you can run a first clip before paying anything — including a low-resolution preview so you can judge the motion and framing. Upgrading buys more credits and higher resolution, not access to the tool itself.
No. The generator runs on managed GPUs and returns a downloadable clip. There is nothing to install, no CUDA setup, and no model download. Only the local ComfyUI workflow requires your own hardware.
The model is built around video continuation, so you generate a base shot and then extend it segment by segment while subject identity, palette, and camera motion carry forward. That is what makes minutes-long, coherent output practical instead of a stitched patchwork.
Output targets 720p at 30fps, with an inexpensive 480p path for testing a prompt. The exact credit cost is shown next to each resolution before you submit a job.
No. This is an independent hosted workspace for the open-weights LongCat Video model. The model authors and the weights live on the upstream GitHub repository; this site provides the browser generator, credits, and documentation around it.
Text-to-video, image-to-video, and continuation in one browser workflow — 720p output with no install and no GPU.