Short answer: in the LongCat generator on this site, one regular (non-distilled) generation can be up to 961 frames. Video length is simply frames divided by frames per second, so 961 frames is about 64 seconds at 15 fps or about 32 seconds at 30 fps. The fast distilled models are capped at 81 frames (about 5.4 seconds at 15 fps). Each generation costs a flat 150–300 credits depending on the model you pick, not on how many frames you ask for. The rest of this guide explains how those numbers fit together and how to plan a longer video without wasting credits.
Why "How Long" Is Really a Frames Question
Most AI video tools quote length in seconds. LongCat is controlled in frames, and seconds fall out of a simple formula:
length in seconds = number of frames ÷ frames per second (fps)
That matters because the same frame count gives very different lengths depending on fps. The generator below lets you set both, so it is worth understanding the trade-off before you press Generate.
| Frames | At 15 fps | At 24 fps | At 30 fps |
|---|---|---|---|
| 81 (distilled maximum) | ~5.4 s | ~3.4 s | ~2.7 s |
| 162 (default for regular models) | ~10.8 s | ~6.8 s | ~5.4 s |
| 480 | 32 s | 20 s | 16 s |
| 961 (regular maximum) | ~64 s | ~40 s | ~32 s |
Lower fps stretches the same frames over more seconds; the generator's fps field accepts 1–60 and starts at 15. Higher fps gives smoother motion but a shorter clip for the same number of frames.
Try It: Set Frames and FPS in the Real Generator
The generator below is the same one on the LongCat home page. Pick text-to-video or image-to-video and look for the frame count and fps fields in the video settings; they are the controls described in this guide.
A practical first run: choose 480p, keep the default 162 frames at 15 fps (about 11 seconds), and write a prompt with one subject and one camera move. Once you like the motion, raise the frame count for the final version.
The Four Video Options and What Each Costs
The generator offers four LongCat video models. The credit cost is a flat amount per generation, taken from the site's pricing logic:
| Option | Resolution / motion | Frame range | Credits per generation |
|---|---|---|---|
| Distilled 480p (fast) | 480p, fewer sampling steps | 17–81 frames | 150 |
| Regular 480p | 480p / 15 fps preset | 17–961 frames | 200 |
| Distilled 720p (fast HD) | 720p, fewer sampling steps | 17–81 frames | 250 |
| Regular 720p | 720p / 30 fps preset | 17–961 frames | 300 |
Two details are easy to miss:
- Asking a distilled model for more than 81 frames switches it to the regular model. The generator does this automatically and the quote changes with it, so a "fast" job with 200 frames is billed as the regular model.
- The cost does not grow with frame count. A 162-frame and a 961-frame regular 480p job cost the same 200 credits. Longer jobs take longer to finish, not more credits — which makes it sensible to test short and then render long. See pricing for the current credit packs.
How Long Does Generation Take?
Rendering time grows with frame count. The generator's own progress estimate starts at about two minutes for the default 162 frames and adds roughly 0.6 seconds of waiting per extra frame, so a 961-frame job is estimated at around 10 minutes. These are estimates shown while you wait, not guarantees; queue load and resolution change the real number. If you are iterating on a prompt, short distilled runs give you feedback in the fewest minutes.
What the Model Itself Is Built For
LongCat-Video is Meituan's open video model. Its public model card describes it as a single model for text-to-video, image-to-video and video continuation, pretrained on continuation so that long outputs keep the same subject, lighting and camera logic instead of drifting after a few seconds. That is the reason this site can offer up to 961 frames in one job when many tools stop at 5 or 10 seconds.
Longer is not automatically better, though. As a rule of thumb, a single, well-defined action is the easiest thing to keep coherent over a long clip, while a prompt that asks for many scene changes in one generation gives the model more chances to blur them. If your story has several scenes, generate them as separate clips and edit them together.
How to Plan a Longer AI Video
- Decide the final length first. For a 30-second social video, 480 frames at 15 fps or 961 frames at 30 fps both work; choose by whether you need smoothness (30 fps) or detail per credit (15 fps).
- Prototype with distilled 480p. At 150 credits and 81 frames, it is the cheapest way to check that the subject, framing and motion match your idea.
- Lock one subject and one camera move. "Slow dolly-in on a barista pouring latte art, warm morning light" stays coherent far longer than a prompt with three actions.
- Start from an image when identity matters. Image-to-video keeps a product or character consistent; the image-to-video extension tutorial covers this workflow.
- Render the long version once. Switch to a regular model, raise frames, and generate the final clip. Because cost is flat per generation, one long render is cheaper than stitching many short ones.
- Edit multi-scene stories outside the generator. Generate each scene as its own clip and join them in an editor.
If you prefer running the open model yourself, our extra-long generation guide for ComfyUI shows how continuation is chained locally, and the long AI video generator page summarises the online option.
LongCat Length Compared With Typical Short-Clip Tools
Many popular video generators are designed around short clips of roughly 5–10 seconds per generation, and longer results are built by extending or stitching clips. LongCat's difference is that a long clip can come out of one generation. That is useful for product turntables, ambient loops, slow establishing shots and talking-style scenes where cuts would be distracting. For fast-cut ads with many scenes, a short-clip workflow plus editing is still a reasonable choice, and you can use LongCat for the shots that need to run longer.
Frequently Asked Questions
What is the maximum length of a LongCat video?
On this site, one regular LongCat generation can be up to 961 frames. That is about 64 seconds at 15 fps or about 32 seconds at 30 fps. Distilled (fast) models are limited to 81 frames.
Does a longer LongCat video cost more credits?
No. Each generation has a flat cost — 150, 200, 250 or 300 credits depending on the model — regardless of frame count. Longer videos take more time to render, not more credits.
Why did my distilled job cost more than expected?
If you request more than 81 frames on a distilled model, the generator automatically uses the regular model instead, and the regular model's price applies.
Should I use 15 fps or 30 fps?
Use 15 fps when you want more seconds from the same frames and the motion is slow. Use 30 fps for smoother fast motion, accepting a shorter clip for the same frame count.
Can I make a multi-minute LongCat video?
Not in a single online generation here; the limit is 961 frames. For multi-minute pieces, generate several long clips and join them in an editor, or chain continuation locally with the open model in ComfyUI.