13.6B dense architecture
All parameters stay active so LongCat Video keeps subjects stable across frames.
LongCat Video is a 13.6B dense model that keeps motion smooth, color stable, and prompts faithful for text-to-video, image-to-video, and video continuation. Generate coherent 720p 30fps sequences that actually run minutes long.
Cost 150 credits
Remaining 0 credits
Preview
Nothing generated yet
LongCat Video is an independent wrapper that provides a user-friendly interface on top of LongCat and other third-party AI models. We focus on clearer UI, extra video tools, flexible pricing, and support that replies within 3 business days.
Everything you need to keep long videos coherent and on-brand.
All parameters stay active so LongCat Video keeps subjects stable across frames.
Generate fast drafts then refine to crisp 720p 30fps without rerunning the whole clip.
Less than 10% of the usual attention cost, so LongCat Video runs longer without melting GPUs.
Balances visual quality, motion quality, and text alignment for prompt-faithful outputs.
Pretrained on continuation, so extending clips feels natural with fewer sudden resets.
Use LongCat Video commercially, fine-tune privately, and deploy the workflow anywhere.
Dense parameters, continuation-native training, and efficient attention keep stories stable across hundreds of frames.
One LongCat Video stack covers prompts, still-frame extensions, and clip stitching so teams can storyboard once and render straight through.
Train-time video continuation plus GRPO rewards keep objects stable and motion natural for multi-minute runs.
All parameters fire every frame, helping LongCat Video hold alignment and visual quality without huge hardware.
MIT licensing, reproducible seeds, and browser-friendly tools make LongCat Video practical for research, product, and creative pipelines.
Open, fast, and consistent for real productions.
13.6B dense
All-parameter LongCat Video core for stable frames
720p 30fps
Default LongCat Video quality for smooth playback
Minutes-long
Continuation + GRPO keep stories coherent past 60s
MIT
Open-source LongCat Video license for commercial use
Prompt or upload â?LongCat Video â?publish. Long clips without jitter.
Describe scene, camera, motion, and duration. Add seed for reproducible LongCat Video runs.
Use image-to-video or continuation to keep style and layout locked while LongCat Video extends the story.
Coarse-to-fine rendering plus Block Sparse Attention keep LongCat Video outputs smooth for minutes. Export MP4, GIF, or thumbnails.
Answers about quality, licensing, hardware, and prompt writing for LongCat Video.
Need help? Email support@longcat-video.comLongCat Video refers to the open-source model family and this site provides an independent, creator-friendly interface for experimenting with those workflows.
Tutorials, guides, and updates about LongCat Video.
Jan 31, 2026
Prompt techniques, resolution strategies, and workflow tips for creating professional sports training videos with AI.
Read moreJan 25, 2026
Learn how to use LongCat AI video generation, ComfyUI workflows, download guides, and advanced prompting techniques.
Read moreJan 24, 2026
Why continuity, motion stability, and prompt fidelity make LongCat Video useful for long-form creative production.
Read moreUse LongCat Video to turn prompts or images into minutes-long 720p 30fps stories without drift.
Start generating with LongCat Video