NewKling 4.0 is live on imageatKling 4.0
30-second takes. 10 keyframes.
Kling AI's fourth-generation video model makes 3 to 30 seconds of video in one generation, steered by up to 10 keyframes and 15 references, with multi-shot prompts, stereo audio and text-guided video editing.
Kling 4.0 at a glance
Kling 4.0 is Kling AI's newest AI video model, released on October 15, 2026. It generates 3–30 second videos from text, a start and end frame, up to 10 keyframes or up to 15 Omni references, at 720p, 1080p or 4K, in 16:9, 9:16, 1:1 or 21:9, with stereo audio. A faster Kling 4.0 Flash makes 3–20 second 720p clips. On the API it costs $0.157 per second at 720p and $0.235 at 1080p; on imageat it is available in the AI video studio at 5 credits per second.
Updated October 21, 2026 · Sources: Kling AI release notes, fal model pages
Kling 4.0 in motion
Samples from Kling's release notes, served from imageat's CDN. These are Kling's demonstrations, not imageat generations.
Motion and camera control
A long take where the subject and the camera both move, the kind of shot Kling 4.0's 30-second single pass is built for.
Reference-guided generation
Omni Reference keeps characters, products and places from your images and clips consistent through the shot.
What's new in Kling 4.0
Kling 4.0 is built for longer, directed shots: more length per generation, more control points along the timeline, and more ways to bring your own characters and footage into the scene.
30-second single takes
Kling 4.0 renders up to 30 seconds in one pass, double Kling 3.0's 15. A camera move or a sequence of actions carries through the whole clip instead of being stitched from short pieces.
Up to 10 keyframes
Beyond a first and last frame, you can pin up to 10 images along the timeline. Kling animates the path between them, which gives you control over character states, scene changes and pacing.
Omni Reference
Mix up to 15 references in one job: images, video clips and saved subjects with front and side views. Call them in the prompt as @Image1, @Video1 or @Element1 to place them in the scene.
Multi-shot prompts
Split a video into shots, each with its own prompt and length. The shots run as one generation, so cuts stay consistent in character, lighting and style.
Stereo audio and lip sync
Kling describes two-channel stereo sound and tighter synchronization between dialogue and the speaker's mouth and performance.
Video editing
Upload a 3 to 30 second clip and describe the change: a new background, weather, outfit or style. Add reference images for the new look. The edit keeps the source length.
Kling lists 10-bit HDR at 1080p and 4K and video extension up to about two minutes as coming soon, and allows prompts of up to 8,000 tokens in its own app.
Kling 4.0 modes
and input limits
The four Kling 4.0 endpoints fal publishes for developers, with the limits each one accepts.
| Mode | Input | Output |
|---|---|---|
| Text to video | A prompt of up to 2,500 characters, or per-shot prompts | 3–30s · 720p/1080p · 16:9, 9:16, 1:1, 21:9 |
| Image to video | A start image, plus an optional end image to land on | 3–30s · frame follows your images |
| Reference to video | Up to 10 images, 5 videos (30s total) and 7 elements, referenced as @Image1, @Video1, @Element1 | 3–30s · Auto ratio follows the references |
| Video editing | A 3–30s source clip (min 700px, 24–60 fps) and an edit instruction, plus optional reference images | Keeps the source length unless you set one |
How to prompt Kling 4.0
Write it as shots
For anything over 10 seconds, split the idea into shots: "Shot 1, 6s: …". Give each a framing, an action and a camera move.
Name your references
Point to each upload by its tag: "@Element1 sits at the table in @Image1 and lifts the cup from @Image2".
Pin the hard moments
Use a last frame or keyframes for the poses you need exactly, and let the prompt describe the motion between them.
Write the sound
Put dialogue in quotes and name the ambience: rain on glass, a crowd, footsteps on gravel.
Kling 4.0 vs Flash vs 3.0
| Capability | Kling 4.0 | Kling 4.0 Flash | Kling 3.0 |
|---|---|---|---|
| Released | October 15, 2026 | Limited early access, September 28, 2026 | 2026 |
| Length | 3–30 seconds | 3–20 seconds | 3–15 seconds |
| Resolution | 720p, 1080p, 4K listed (10-bit HDR coming soon) | 720p, 8-bit SDR | 720p, 1080p, 4K |
| Frames | First, last and up to 10 keyframes | First frame | First and last frame |
| References | Omni Reference, up to 15 items | Omni Reference | Up to 4 images |
| Editing | Text-guided video editing | Not listed | Motion control |
| Aspect ratios | 16:9, 9:16, 1:1, 21:9 | Not specified | 16:9, 9:16, 1:1 |
Kling 4.0 and Flash rows are Kling's published specifications; Kling 3.0 is as offered on imageat. Availability can differ by mode during the rollout.
Kling 4.0 timeline
Sep 28, 2026
Kling 4.0 Flash opens to a limited group of members.
Sep 30, 2026
Kling announces early access to Kling 4.0 and publishes the release notes.
Oct 15, 2026
Kling 4.0 officially launches on klingai.com for all users.
Oct 2026
fal lists four Kling 4.0 API endpoints; developer access is marked coming soon.
Kling 4.0 API
fal publishes four endpoints. They opened to API callers on October 24, 2026, and imageat runs Kling 4.0 on them.
kling/v4/text-to-video
kling/v4/image-to-video # start_image_url, end_image_url
kling/v4/reference-to-video # image_urls, video_urls, elements
kling/v4/video-to-video/edit # video_url, image_urls
{ "prompt": "...", "duration": 10,
"resolution": "1080p", "aspect_ratio": "16:9" }On imageat, pass kling4 to the video API or MCP server, one key for every model.
Kling 4.0 price per second
On imageat, a 5-second 720p clip is 25 credits and a 10-second 1080p clip is 70. Editing bills each second of the source clip. One credit is $0.10, and failed generations are refunded automatically.
fal list price
- Text, image, reference images
- $0.157/s 720p · $0.235/s 1080p
- Editing, video references
- $0.094/s 720p · $0.141/s 1080p, on input + output (min 2× output)
Make your first
Kling 4.0 video
Kling 4.0 sits in the imageat video studio next to Seedance, Wan and Veo. Write the shot, pick a length, and pay only for the seconds you generate.
Kling 4.0 questions
What is Kling 4.0?
+
Kling 4.0 is the fourth-generation AI video model from Kling AI, the video team at Kuaishou. It makes 3 to 30 seconds of video in one generation from text, images, keyframes or references, with stereo audio. Its headline additions over Kling 3.0 are 30-second single takes, up to 10 keyframes per clip, Omni Reference with up to 15 images, videos and subjects, and multi-shot prompts.
When was Kling 4.0 released?
+
Kling 4.0 Flash opened to a limited group of members on September 28, 2026. The full Kling 4.0 model officially launched on klingai.com on October 15, 2026. The developer API was listed as coming soon at launch; fal opened its Kling 4.0 endpoints to API callers on October 24, 2026.
How long can a Kling 4.0 video be?
+
From 3 to 30 seconds in a single pass, chosen one second at a time. With multi-shot prompts the clip is split into shots that each get their own prompt and length, and the shot lengths add up to the video length. Kling says extending a video up to about two minutes is coming later.
What resolution does Kling 4.0 support?
+
Kling lists 720p, 1080p and 4K, with 10-bit HDR at 1080p and 4K marked as coming soon. The fal API endpoints offer 720p and 1080p. Aspect ratios are 16:9, 9:16, 1:1 and the new 21:9 ultrawide; image-to-video follows the shape of your input frames.
What is the difference between Kling 4.0 and Kling 4.0 Flash?
+
Flash is the faster, lighter variant: 3 to 20 second videos at 720p in 8-bit SDR, from text, a first-frame image or Omni Reference. The full Kling 4.0 model goes to 30 seconds and up to 4K, and adds first-and-last frame, up to 10 keyframes and video editing.
How many references can Kling 4.0 use?
+
Kling says Omni Reference accepts up to 15 assets in one job, within per-type limits. On the fal API that means up to 10 reference images, 5 reference videos of 3 to 30 seconds (30 seconds combined), and 7 elements, which are characters or objects built from a front view plus extra angles. Up to 3 elements can carry a voice. In the prompt you call them @Image1, @Video1 and @Element1.
Can Kling 4.0 edit an existing video?
+
Yes. Kling 4.0 Video Editing takes a 3 to 30 second clip and a text instruction, such as changing the background, the weather or an outfit, and can use reference images for the new look. The output keeps the source video's length unless you set another duration. The source must be at least 700 pixels on each side and 24 to 60 fps.
Does Kling 4.0 generate sound?
+
Kling's release notes describe two-channel stereo audio and improved lip sync between speech and on-screen performance. Write the dialogue in quotes and describe the ambience and sound effects in the prompt.
How much does Kling 4.0 cost?
+
On fal, Kling 4.0 is billed per second of output: $0.157 at 720p and $0.235 at 1080p. Video editing and runs with video references cost $0.094 per second at 720p and $0.141 at 1080p, but count the input plus the output, and at least twice the output length. On imageat, Kling 4.0 costs 5 credits per second at 720p and 7 at 1080p; editing costs 5 or 8 credits per second of source video. Failed generations are refunded automatically.
Is there a Kling 4.0 API?
+
Yes. fal serves four Kling 4.0 endpoints, kling/v4/text-to-video, image-to-video, reference-to-video and video-to-video/edit, open to API callers since October 24, 2026. On imageat, use the model id kling4 in the video API or MCP server, with one key across all models, from $0.32 per second at 720p.
Can I use Kling 4.0 on imageat?
+
Yes. Pick Kling 4.0 in the AI video studio to generate from text or an image, or in the AI video generator for first-and-last frame and reference images. Video editing is in the studio's Edit tab. You pay per second, and failed generations are refunded automatically.
Kling is developed by Kuaishou. imageat is not affiliated with Kuaishou or Kling AI.