What is Kling 4.0?
Kling 4.0 is Kuaishou's fourth-generation Kling AI video model, announced at the end of September 2026 with a full launch set for October 2026. It is not yet callable on Unifically. We will open the Kling 4.0 API here once Kuaishou serves it, and every spec on this page comes from the release and will be re-verified against the callable API on day one.
The model generates 3 to 30 second videos at 720p, 1080p, or 4K from a prompt, a start image, a first and last frame pair, up to 10 keyframes, or a bundle of up to 15 reference assets. Audio renders in the same pass as two-channel stereo, with lip-synced dialogue in more than nine languages. Prompts run up to 8,000 tokens, and short prompts get expanded into fuller shot direction before generation. The same model also edits existing videos, so generation, references, and editing live in one place.
Key features of Kling 4.0
- 30 seconds in one generation. Any length from 3 to 30 seconds, enough for a long take or a complete short scene without joining separate videos.
- Up to 10 keyframes. Pin the character states, scene changes, and story beats that matter, and the model generates the motion between them. First and last frame control stays available.
- Omni Reference with 15 assets. Combine up to 10 images, 5 videos (30 seconds combined), 7 subjects, and a voice reference in one generation. Image references can be character sheets, storyboard grids, wireframes, or game UI; video references can be plain motion, white-model renders, or depth maps.
- Stereo audio with lip-sync. Dialogue, effects, and ambience render together in two-channel stereo, with accents and dialects such as Cantonese and British English.
- 720p to 4K. Resolutions of 720p, 1080p, and 4K in 21:9, 16:9, 1:1, 9:16, or auto.
- Video editing and recreation. Change expressions, movement, camera position, style, weather, or background in up to 5 input videos while the rest of the scene holds, or copy the pacing and shots of an existing video with your own product or characters swapped in.
- Readable on-screen text. Signs, packaging, and captions render legibly in nine-plus languages and stay stable while the camera moves.
Best for
Long single takes
A 30-second ceiling covers a full vlog segment, ad, or story beat in one pass.
Storyboard-driven scenes
Up to 10 keyframes let you lock the story beats and let the model fill the motion between them.
Multi-character productions
Up to 7 subjects and 10 reference images keep a cast, props, and vehicles consistent across one video.
Dialogue in many languages
Lip-synced speech with regional accents renders with the picture, so a talking scene comes back finished.
Editing instead of regenerating
Swap a character, background, or style in an existing video and keep the performance and timing.
Remaking proven ads
Creative recreation copies the pacing and shots of a reference video and swaps in your own product or cast.
Use cases
Turn a single reference photo into a 30-second vertical vlog that moves between shots without stitching. Lay out a product launch as keyframes at the moments that matter and let the model build the transitions. Feed a cast sheet, a car, and a set of props into one generation to produce a short action scene where every character stays on model. Localize a talking-head ad by swapping the presenter and the set while the script, timing, and subtitles stay put, or regenerate the dialogue in Spanish, Hindi, or Cantonese with matching lip-sync. And take a trending format, keep its pacing and camera work, and recreate it with your own brand assets instead of starting from a blank prompt.
Limitations
Kling 4.0 is not on Kuaishou's developer API yet, so request fields, durations per resolution, and prices are unknown until it opens. 10-bit HDR at 1080p and 4K is announced as coming soon, and video extension to 2 minutes is not released either. Kuaishou has published no generation times, frame rates, or credit costs for the model so far. No public leaderboard lists Kling 4.0 as of October 1, 2026, so its quality against Seedance, Veo, and FLUX 3 Video is not yet independently measured.
Kling 4.0 vs Kling 3.0
Kling 4.0 doubles the single-generation ceiling from Kling 3.0's 15 seconds to 30 seconds. Kling 3.0 controls only the start and end frames; 4.0 adds up to 10 keyframes in between. References grow the most: Kling 3.0 Omni shares a limit of 7 images and subjects, or 1 video plus 4 images when a video is involved, while 4.0 takes 10 images, 5 videos, and 7 subjects in one bundle of up to 15. 4.0 also adds a voice reference, stereo audio, 21:9 output, Portuguese, German, French, and Hindi dialogue, and 8,000-token prompts.
Kling 3.0 stays the practical choice on Unifically today: it is live from $0.06 per second with 4K output, multi-shot mode, and native audio. Once 4.0 lands, long takes, keyframe control, and multi-character casts are the reasons to switch.
How to get access to Kling 4.0
Kling 4.0 Flash is open now to a limited group of users inside the Kling app. The full Kling 4.0 model launches in October 2026, and neither model is on Kuaishou's developer API yet. When the API opens, Kling 4.0 will appear on Unifically with a browser playground on this page, a REST endpoint under the same API key as every other model, and pay-per-second billing from one balance. Until then, Kling 3.0 and Kling 3.0 Omni cover text-to-video, image-to-video, and reference-driven generation.
What Kling 4.0 can do
30 seconds from one reference photo
One reference image of a man in a gym becomes a 30-second vertical vlog in a single generation: he talks to the camera, presses dumbbells, drinks water, and talks again, with his face, glasses, and the room holding steady across every shot change.
Five keyframes, one continuous video
Five keyframes of glowing light drawings in a dark bedroom (a hand, a heart on a monitor, an octopus, a burger in the fridge, an eye on the wall) become one 21-second video that moves through each beat in order. This is the control Kling 3.0 lacks: you set the moments between the first and last frame.
A full cast from one reference sheet, in 4K
One reference sheet with three actors, a motorcycle, a race car, and a film crew rig becomes a 15-second 4K chase: a driver at the wheel, a cameraman on the bike filming the car, and a stunt landing in front of a fireball. Each person and vehicle keeps its look from the sheet.
Swap the presenter and the set, keep the take
Video editing with one reference image: a woman talking on a grey sofa becomes a different woman in a red dress in a white Mediterranean living room. Her line, gestures, timing, and the burned-in subtitles carry over from the source video unchanged.
Dialogue, footsteps, and street sound in one pass
A handheld 7-second reunion on a British street at dusk, with English dialogue, footsteps, street ambience, and music generated together with the picture. The two friends cross the road, laugh, and hug with natural timing.
A six-person fight that stays readable
Fight choreography with several people is a known weak spot for video models. Here one woman takes down five men in turn across a hotel lobby, and each strike, fall, and turn stays clear across 15 seconds.
FAQs
People also ask
Kling 4.0 is Kuaishou's fourth-generation Kling AI video model. It generates 3 to 30 second videos at 720p, 1080p, or 4K from text, images, up to 10 keyframes, or a mix of up to 15 reference assets, with two-channel stereo audio and lip-synced dialogue. It also edits existing videos.
Kling 4.0 is the latest Kling video model. Kuaishou announced it at the end of September 2026, opened the faster Kling 4.0 Flash to a limited group of users on September 28, 2026, and set the full Kling 4.0 launch for October 2026. Kling 3.0 is the newest Kling generation you can call through an API today.
Kuaishou set the full Kling 4.0 launch for October 2026 without naming a day. Kling 4.0 Flash, the fast model in the same lineup, has been in limited early access since September 28, 2026.
Not yet. Kuaishou's developer API does not serve Kling 4.0 so far, and its current API plans exclude it. Unifically will open Kling 4.0 once the API is live, and this page will get a browser playground, callable parameters, and per-second pricing at that point.
Kuaishou has not published Kling 4.0 credit or API prices yet. Unifically will bill it per second of output with no subscription, like every Kling model here, and will post the rate on this page at launch. For comparison, Kling 3.0 on Unifically starts at $0.06 per second.
There is no free Kling 4.0 API. Every new Unifically account gets $0.50 of free balance with no subscription, which you can spend in the playground on Kling 3.0 and other live models today, and on Kling 4.0 once it launches here.
3 to 30 seconds in a single generation, double the 15-second ceiling of Kling 3.0. A video extension feature that grows a video forward or backward to 2 minutes in total is announced as coming soon.
720p, 1080p, or 4K, in 21:9, 16:9, 1:1, 9:16, or auto aspect ratios. 10-bit HDR output at 1080p and 4K is announced as coming soon rather than available at launch.
Up to 15 assets in one generation, combining up to 10 images, up to 5 videos with 30 seconds of combined length, up to 7 subjects (3 of them video-based), and a voice reference. Image references can be character sheets, storyboard grids, or wireframes, and video references can be white-model renders or depth maps.
Dialogue covers Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, Indonesian, Hindi, and more, plus accents and dialects such as Cantonese, Sichuanese, and British, American, and Indian English. On-screen text renders in the same languages, along with emoji.
Kling 4.0 is the full model, with 3 to 30 second videos up to 4K, first and last frames, up to 10 keyframes, and video editing. Kling 4.0 Flash is the fast, lower-cost model, with 3 to 20 second videos at 720p and text-to-video, image-to-video, and Omni Reference only.
