If you typed long cat video, you are looking for LongCat Video: Meituan's open-weights video model, written with a space because that is how the name reads out loud. There is no separate "long cat" model — the space, the hyphen (longcat-video) and the run-together form (longcatvideo) all point at the same 13.6B-parameter model.
LongCat is the model's proper name; long cat video is the same words with the camel case flattened.meituan-longcat/LongCat-Video, so GitHub searches often land on the hyphenated form.LongCat Video takes text, images, or an existing clip and produces motion:
Hosted output targets 720p at 30fps; a cheaper 480p path lets you test a prompt before committing.
The hosted generator runs the model on managed GPUs and hands back a downloadable file. If you would rather run it yourself, the local setup guide lists the commands and VRAM figures; if you want the model's history and its sibling models, see LongCat AI.
Yes. They are the same model. "Long cat video" is simply the space-separated spelling of LongCat Video, and the hyphenated form longcat-video in the upstream GitHub repository refers to the same thing.
No. LongCat Flash is a language model for text reasoning. LongCat Video is the video generation model that produces text-to-video, image-to-video, and continuation output.
Yes. The generator on this site runs the model on managed GPUs in the browser, so you can produce a clip with no local install, no CUDA setup, and no model download.
LongCat Video supports continuation, so rather than generating one short clip you extend a scene segment by segment while the subject and look stay consistent. That is what allows minutes-long results.
The open-weights model and its documentation live in the upstream Meituan LongCat Video repository. This site is an independent hosted workspace around that model, plus documentation and a browser generator.
The same LongCat Video model, hosted — text-to-video, image-to-video, and continuation with no install.