3 to 20 seconds per run
Set the duration anywhere from 3 to 20 seconds in a single generation, long enough for a full social clip or ad cut without stitching.
KL AI 4.0 Flash is the fast, lower-cost member of the 4.0 video generation. It turns a prompt, a single first frame or a set of references into 3 to 20 second clips at 720p, and costs about 40% less per second than the full KL AI 4.0.
Flash belongs to the same 4.0 generation as the full KL AI 4.0 and shares its prompt budget, reference limits and request types. It is a speed and price tier: it stops at 720p and 20 seconds and drops multi-keyframe control, in exchange for faster turnaround and a lower rate per second. That makes it a drafting and volume tool rather than a mastering tool.
September 28, 2026
The 4.0 generation is announced. Flash goes out first, as early access for top-tier annual subscribers only.
October 28, 2026
The full KL AI 4.0 becomes generally available through the API: up to 30 seconds, 720p and 1080p, up to 10 keyframes.
November 4, 2026
Flash opens to every user and gets a public API with a published per-second price. The waitlist and the annual-plan requirement end.
Set the duration anywhere from 3 to 20 seconds in a single generation, long enough for a full social clip or ad cut without stitching.
Flash renders at 720p in 8-bit SDR. That suits feeds, vertical ad slots and edit previews; there is no 1080p or 4K tier for Flash.
Start from a prompt, from one first-frame image, or from an Omni reference task with up to 10 images and 5 videos, 15 inputs in total.
Prompts can run to 26,000 English characters, and the aspect ratio can be 21:9 or adaptive in addition to the usual 16:9, 9:16 and 1:1.
The two share a generation, a prompt budget and a reference budget. The differences are length, resolution, keyframes and price.
| Feature | KL AI 4.0 Flash | KL AI 4.0 |
|---|---|---|
| Clip length | 3–20 seconds | 3–30 seconds |
| Resolution | 720p only, 8-bit SDR | 720p or 1080p; 10-bit 4K announced |
| Image-to-video control | One first-frame image | Up to 10 keyframes along the timeline |
| References | Up to 10 images, 5 videos, 15 inputs | Same limits, plus direct voice references |
| Price per second at 720p | About 40% lower | Baseline; 1080p costs about 50% more than 720p |
Prices compare the developer's published API rates for runs without input video. Credits on kling3.org are shown in the studio before you generate.
Flash and the full model read prompts the same way, so you can write and refine your 4.0 prompts in the studio right now.
The button on this page opens the video studio with KL AI 4.0 selected. Sign in to see your credit balance.
Write the subject, action, camera move and mood, add reference images if you have them, and pick 3 to 30 seconds at 720p or 1080p.
Short 720p drafts keep costs low, which is the Flash way of working. Once the motion is right, render the final take at 1080p.
It is the fast, lower-cost tier of the 4.0 video generation. It makes 3 to 20 second clips at 720p from text, a single first frame or Omni references, at about 40% less per second than the full KL AI 4.0.
Since November 4, 2026, anyone. Before that, from September 28, it was limited to top-tier annual subscribers as early access.
No. Flash outputs 720p only and its image-to-video mode takes a single first frame. For 1080p, clips up to 30 seconds or up to 10 keyframes, use the full KL AI 4.0.
On the developer's API rate card, a second of 720p Flash output without input video costs about 40% less than the same second from the full model. With input video, both tiers bill on a longer total duration, so check the full cost before relying on the lower rate.
It is the same generation with the same reference and prompt limits. What it gives up is resolution, length and keyframe control, so it fits drafts and 720p delivery better than final masters.
The full KL AI 4.0 is ready in the studio: up to 30 seconds, 720p or 1080p, from a prompt or your own images.