Models / Kling AI / Kling 4.0
Kling 4.0 AI Video Generator
Kling 4.0 by Kuaishou is an AI video model that makes 3 to 30 second clips at up to 4K, with two-channel stereo sound, up to 10 keyframes and up to 15 references in a single generation.
Coming soon on AgentDocks
- Made by
- Kuaishou
- Length
- 3 to 30 s
- Resolution
- 720p to 4K
- Sound
- Two-channel stereo
- Keyframes
- Up to 10
- References
- Up to 15
Overview
What is Kling 4.0?
Kling 4.0, also called Kling 4, is the fourth generation of Kling AI, the video model family from Kuaishou. It turns a prompt, images or reference videos into clips of 3 to 30 seconds at 720p, 1080p or 4K, with stereo sound and lip-synced dialogue in many languages and accents. Kuaishou built it around three things: more realistic motion and faces, finer control through keyframes and references, and longer stories told in one take. A lighter version, Kling 4.0 Flash, makes 3 to 20 second clips at 720p for fast, frequent work.
Real motion and faces
Fast action and complex camera moves stay smooth, and faces, skin and light hold up in close-up, down to a glance or a moment of hesitation.
Control with keyframes and references
Up to 10 keyframes and 15 references set where the shot goes, who is in it and what stays the same.
Longer stories in one take
Up to 30 seconds in a single generation, with dialogue, sound and story beats that hold together.
What's New
What's New in Kling 4.0
01
Clips up to 30 seconds
One generation runs up to 30 seconds, double Kling 3.0's 15, so a scene can play out as one continuous take instead of clips cut together.
02
Up to 10 keyframes
Place up to 10 keyframe images along the clip to set where characters are, what changes and when. Kling 4.0 fills in the motion between them.
Kling 4.0 prompt
A 12-second product film for loopa sparkling kombucha. From the overhead shot on crushed ice, the camera rises as the bottle lifts upright and yuzu, ginger and ice burst around it in slow motion. The bottle then tilts and pours into a tall glass of ice with rising bubbles, ending on a clean pack shot of the bottle and the glass. Smooth, high-speed product camera work, crisp light, label readable throughout. Sound: ice crackle, a cap hiss, a fizzing pour, a bright pop beat ending on a bass hit.
The stills · 4 image prompts
Keyframe 1 · GPT Image 2.5 Sunburst
An overhead beverage campaign photograph in a wide 16:9 frame, camera pointing straight down. A glass bottle of sparkling kombucha lies on its back in the center, label facing up and fully readable: a bright yellow-green label with playful teal and orange shapes, the word "loopa" in a rounded hand-drawn logo, and "GINGER + YUZU" and "SPARKLING KOMBUCHA 330 mL" beneath. Silver cap toward the top. Irregular clear crushed ice fills the frame around it on a deep clean teal base, with whole yuzu, a halved yuzu, two wedges and two small ginger slices around the edges, some cropped. Crisp diffused studio light from the upper left, wet ice facets, small meltwater drops.
Keyframe 2 (keyframe 1 attached) · GPT Image 2.5 Sunburst
The same loopa bottle from the attached image, now upright and centered against a deep teal background in a wide 16:9 frame, with yuzu halves, wedges, ginger slices and chips of ice bursting outward around it in mid-air, frozen at high speed, droplets sparkling. Label facing the camera and readable.
Keyframe 3 (keyframe 1 attached) · GPT Image 2.5 Sunburst
The same loopa bottle from the attached image, tilted and pouring sparkling golden kombucha into a tall glass full of ice on a teal surface, in a wide 16:9 frame. Bubbles rising, a yuzu wheel on the rim, splashes frozen in mid-air, label readable.
Keyframe 4 (keyframe 1 attached) · GPT Image 2.5 Sunburst
A final pack shot in a wide 16:9 frame: the loopa bottle from the attached image and a full glass of iced kombucha side by side on a teal surface, with a few ice chips and a halved yuzu, condensation on the glass and clean space above. Label square to the camera and readable.
Stills made with GPT Image 2.5 Sunburst on AgentDocks, based on the Drink on Crushed Ice template.
What Went In
What Came Out
03
15 references in one run
Mix up to 10 images, 5 videos (30 seconds in total) and 7 subjects, including a subject's voice, to keep characters, products and style steady across the clip.
Kling 4.0 prompt
The woman from the character image stands at the yellow tiled counter in the bathroom from the location image, holding the rinzo bottle from the product image up beside her face, label to the camera. She smiles and says, "Okay, it smells like a holiday. I'm obsessed," then pumps a swirl of foam into her palm. Handheld, phone-style vertical framing, natural morning light. Sound: soft room tone, a pump click, a little laugh, light upbeat music.
The stills · 3 image prompts
Character · Nano Banana Pro
A square studio portrait of a woman in her late twenties with voluminous golden-blonde curls, fair skin with warm peach undertones and light freckles, in a white tank top, with a warm natural smile, against a plain light-grey background. Soft daylight, looking at the camera.
Product · Nano Banana Pro
A square product photo of a tall pump bottle of body wash in glossy cobalt blue with a bright orange pump. Its illustrated label shows coral and orange hibiscus flowers, the word "rinzo" in bold white rounded letters, and "CITRUS CURRENT" and "DAILY BODY WASH 400 mL" beneath. Plain white background, front view, label readable.
Location · Nano Banana Pro
A bright bathroom in a wide 16:9 frame: glossy sunflower-yellow square tiles with fine ivory grout, a tiled yellow counter, a sage-green terry towel hanging on the left and a small ivory-framed window at the upper right with soft green foliage outside. Morning sun, a few soap bubbles on the counter. No products, no people.
Stills made with Nano Banana Pro on AgentDocks, based on the Beauty Campaign Portrait template.
What Went In
What Came Out
04
Recreate a video with your assets
Use an existing video as the reference for pacing, shots and camera work, then swap in your own product or character. Kuaishou names product ads, fashion and beauty as the main uses.
05
Edit a video you already have
Change expressions, movement, camera, style or background in an existing clip and keep the rest steady. It takes up to 5 clips, one of them the main video, 30 seconds in total.
Kling 4.0 prompt
Keep the woman, her walk, her timing and the camera move exactly the same. Change the colorway for autumn: the mural becomes rust orange, olive green, mustard and deep brown, and her knit coat and scarf change to the same rust, olive and mustard pattern. Keep the warm light and the real knit texture.
First frame based on the Walking Mural Fashion template.
06
4K, 21:9 and 10-bit HDR
Outputs 720p, 1080p or 4K in 21:9, 16:9, 1:1, 9:16 or Auto. Kuaishou lists 10-bit HDR at 1080p and 4K as coming soon, for more detail in night and backlit shots.
07
Stereo sound and lip sync
Makes two-channel stereo with closer lip sync. Characters speak Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, Indonesian and more, in accents such as American, British and Indian English.
Kling 4.0 prompt
The woman on the left shakes the bag of Savory Lane crackers and says in a British accent, "Last one's mine." The man on the right grabs for it and answers in Indian English, "You said that about the last bag." She laughs and pulls the bag away as a scooter passes behind the car from left to right. Gentle handheld camera, palm shadows moving across the paint. Sound: the crinkle of the bag, a crunch, the scooter passing from left to right, distant traffic, light surf-rock guitar.
First frame · the image prompt
First frame · Nano Banana Pro
A bold lifestyle campaign shot in a wide 16:9 frame, taken from a low angle beside a vintage marigold-yellow convertible on a sunny palm-lined street under a bright blue sky with soft white clouds. Two friends sit in the front seats. On the left, a woman in her twenties with dark wavy hair and big sunglasses holds a bag of cheddar crackers up above the door, label to the camera: an orange bag reading "Savory Lane" and "Cheddar Cheese Rounds" with a photo of golden crackers. On the right, a man in his twenties in a white linen shirt laughs and reaches for the bag. Crisp midday sun, saturated colors, glossy paintwork.
First frame made with Nano Banana Pro on AgentDocks, based on the Snack Car Shot template.
08
Readable text in the video
Writes clear text, emoji and logos in Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French and Indonesian, and keeps them readable while the camera moves or the scene changes.
First frame · the image prompt
First frame · Nano Banana Pro
A playful bakery-counter campaign photo in a wide 16:9 frame. An open paperboard tray, turquoise with a scalloped edge, the word "crumi" in bold coral letters and "COOKIE + SOFT SERVE" beneath, holds a large chocolate-chunk cookie and a cup of vanilla soft serve with cookie crumbs. It sits centered on a smooth marigold-yellow counter in front of a wall of ivory tiles with a yellow tile column on the right, a small chrome service bell at the far right, and an unlit neon sign on the wall that reads "crumi". Bright clean daylight, label square to the camera.
First frame made with Nano Banana Pro on AgentDocks, based on the Cookie Counter Moment template.
Lip Sync
Kling 4.0 Lip Sync Across Languages
One chef, one first frame and the same line in each language. Turn the sound on to hear how the lips follow every word.
English
This is our signature dish, made fresh every morning.
Spanish
Este es nuestro plato estrella, preparado fresco cada mañana.
Japanese
当店の看板料理です。毎朝、新鮮に作っています。
Kling 4.0 prompt · the same prompt for each, with the line swapped
The chef holds the plate toward the camera, smiles and says in English, "This is our signature dish, made fresh every morning." Steam rises from the rice. Small, natural head movement, his lips matching every word. Sound: kitchen clatter, a sizzling pan.
First frame · the image prompt
First frame · Nano Banana Pro
A polished food-advertising portrait in a vertical 9:16 frame. A smiling Middle Eastern chef in his forties with a short grey beard and a crisp white chef's jacket stands in a rich red kitchen, holding a warm-white plate of fluffy long-grain rice with diced carrots, green peas and parsley toward the camera, a thread of steam rising. Behind him are a red enamel casserole, a tall red pepper mill and, on the counter, a bag of rice with a bold label reading "GRAINO" and "LONG GRAIN RICE". Warm saturated light, appetizing food, natural skin.
First frame made with Nano Banana Pro on AgentDocks, based on the Red Kitchen Rice template.
Same Prompt
Kling 4.0 vs Kling 3.0, Side by Side
Each pair ran the same prompt and settings on both models, first take of each, no retries. Kling 3.0 runs on AgentDocks today.
Fast motion
Prompt · the same prompt · 16:9 · 8 s · 1080p · sound on
A fast train blasts past behind her from left to right, its windows streaking into strong horizontal motion blur. She stays almost perfectly still, calm and centered, holding the bouquet steady while the wind from the train whips her hair across her face and tugs at her jacket and the petals. The camera slowly zooms in toward her face. Sound: the roar and rhythmic clatter of the train, a rush of wind, a soft piano note as it passes.
Specs
Kling 4.0 Specs Compared
| Spec | Kling 4.0 | Kling 4.0 Flash | Kling 3.0 |
|---|---|---|---|
| Length | 3 to 30 s | 3 to 20 s | 3 to 15 s |
| Resolution | 720p, 1080p, 4K | 720p | 720p, 1080p, 4K |
| Color | 10-bit HDR at 1080p and 4K (coming soon) | 8-bit SDR | Standard (SDR) |
| Aspect ratios | 21:9, 16:9, 1:1, 9:16, Auto | Not listed | 16:9, 9:16, 1:1 |
| Text to video | Yes | Yes | Yes |
| Image to video | Yes, with first and last frames | Yes | Yes, with first and last frames |
| Keyframes | Up to 10 | Not listed | First and last frame |
| References | Up to 15: images, videos and subjects | Omni Reference | Up to 2 images (more on Kling 3.0 Omni) |
| Sound | Two-channel stereo | Not listed | Native audio |
| Prompt length | Up to 8,000 tokens | Not listed | Up to 3,072 characters on AgentDocks |
| Edit an existing video | Up to 5 clips, 30 s in total | Not listed | Not listed |
Source: Kuaishou's Kling 4.0 announcement.
Prompt Tips
How to Get the Best From Kling 4.0
- 1
Make the first frame first
Build the opening still with an image model you can steer, such as GPT Image 2, until the product, label, set and casting are exactly right, then animate it. Kling 4.0 keeps what the first frame shows, so the label you approved is the label you get.
- 2
Write a long clip as timed beats
For a 20 or 30 second take, split the prompt into beats with times, like the trattoria take above: 0 to 8 s, the chef plates the pasta; 8 to 16 s, the server carries it across the room; 16 to 30 s, the friends dig in. One beat per idea keeps a long clip on track.
- 3
Save keyframes for the moments that must land
Set keyframes where the image has to be exact, such as the product reveal or the final pose, and let the model move freely between them. A few well-placed keyframes usually give smoother motion than many.
- 4
Put dialogue in quotes and name the voice
Write spoken lines in quotation marks, say who speaks, and add the language or accent, for example a British accent or Indian English. Describe the sound around them too: street noise, music, footsteps.
- 5
Give each reference one job
When you mix references, say what each one is for: this image is the character, this video is the camera move, this one is the product. Clear roles stop the result from blending them together.
- 6
Name the camera move
Use plain camera language: push-in, pull-out, pan, tracking shot, orbit. Naming the move gets you closer than describing the feeling of the shot.
- 7
Draft short, then go long
Try the idea as a short clip first, and stretch it to 20 or 30 seconds once the look is right. A long take costs more, so get the shot right before you scale it.
Questions
Kling 4.0, Answered.
Updated 17 October 2026
Kling 4.0 is Kuaishou's AI video model. It turns a prompt, images or reference videos into clips of 3 to 30 seconds at 720p, 1080p or 4K, with two-channel stereo sound, up to 10 keyframes and up to 15 references in a single generation.
Kling 4.0 is made by Kuaishou, the company behind the Kling AI app. AgentDocks is not affiliated with Kuaishou; it runs Kling models next to other leading video models.
Kling 4.0 makes clips of 3 to 30 seconds in one generation, double the 15 seconds of Kling 3.0. Kling 4.0 Flash makes 3 to 20 seconds. Kuaishou also lists video extension up to 2 minutes as coming soon.
Kling 4.0 outputs 720p, 1080p or 4K, in 21:9, 16:9, 1:1, 9:16 or Auto aspect ratios. Kuaishou lists 10-bit HDR at 1080p and 4K as coming soon. Kling 4.0 Flash outputs 720p in 8-bit SDR.
Yes. Kling 4.0 generates two-channel stereo sound with closer lip sync. Characters can speak Chinese, English, Japanese, Korean, Spanish, Portuguese, German, French, Indonesian and more, with accents such as American, British and Indian English.
Keyframes are images you place along the clip to fix what the video shows at that moment, such as a pose or a reveal. Kling 4.0 takes up to 10 and fills in the motion between them; Kling 3.0 takes a first and a last frame.
Up to 15 in one generation: up to 10 images, 5 videos with up to 30 seconds in total, and 7 subjects, of which up to 3 can be based on video. A subject can also carry a voice reference.
Yes. Kling 4.0 can change expressions, movement, camera, style or background in an existing clip while keeping the rest steady. It takes up to 5 clips, one of them the main video, each 3 to 30 seconds and 30 seconds in total.
Kling 4.0 Flash is the faster, lighter version for frequent work. It makes 3 to 20 second clips at 720p in 8-bit SDR, from text, an image or Omni references. The full Kling 4.0 goes to 30 seconds and 4K, with up to 10 keyframes.
Kling 4.0 doubles the longest clip from 15 to 30 seconds, adds up to 10 keyframes where Kling 3.0 takes a first and last frame, takes up to 15 references, adds stereo sound, the 21:9 ratio and prompts of up to 8,000 tokens, and can edit an existing video.
Kuaishou announced Kling 4.0 on 28 September 2026, opened the lighter Kling 4.0 Flash to early users the same day, and made Kling 4.0 available in its own Kling AI app in October 2026. Kling 4.0 is live in Kuaishou's own Kling AI app. It is not in Kuaishou's developer API yet, which is how platforms like AgentDocks run it, so it is coming to AgentDocks as soon as the API opens, and this page will show its price and the button to run it that day. Kling 3.0 runs on AgentDocks today, from text or images, at up to 4K with sound.
Also on the Canvas