Kling 4.0 Is Coming to Flixly: 30-Second AI Video, 4K and Omni Reference

Kling AI has announced Kling 4.0, the next generation of its video model, and it is coming to Flixly. Kling 4.0 makes clips up to 30 seconds long in a single generation, in 720p, 1080p or 4K, with two-channel stereo sound, more accurate lip sync and up to 10 keyframes to steer the story. Its Omni Reference mode takes up to 15 images, videos and voices at once.

Kling 4.0 is not in Flixly's model picker yet. We are preparing it now and will update this post the day it goes live. Until then you can keep creating with Kling 3.0 and the other models in the AI Video Generator.

Every clip below is an example Kling AI published with its Kling 4.0 announcement. We re-encoded them for the web; nothing else was changed.

What is new in Kling 4.0

Kling also announced 10-bit HDR output at 1080p and 4K, and extending a video several times up to two minutes. Kling lists both as coming soon on its side.

There is also a lighter Kling 4.0 Flash, built for quick, frequent drafts: text to video, image to video and Omni Reference, 3 to 20 seconds, at 720p.

Big motion that holds together

Fast action, tracking shots and orbiting cameras are where video models usually fall apart. Kling's examples keep large scenes stable while a lot is moving at once. Here a pianist plays on the bow of a ship in a storm, and the sequence cuts from an aerial shot through the spray to his hands on the wet keys and his face, without losing the scene.

Sound and lip sync

Kling 4.0 generates the soundtrack in the same pass as the picture, now in stereo. Kling says the lip sync is more accurate, so singing and dialogue line up with the mouth. Turn the sound on for this 30-second studio take.

Faces that act

Kling's other headline is realism in the small things: skin, light, and expressions that change mid-sentence. This vlog came from one photo and one line of prompt: "A vlogger-style video in which the woman cheerfully introduces herself to the audience. Maintain a natural, authentic look and realistic image quality."

4K and HDR

Kling 4.0 renders up to 4K, and adds a 21:9 ultrawide format for cinematic framing. Kling's 10-bit HDR output is meant to keep more detail in backlit, night and neon scenes. The example below is from Kling's HDR demo; your browser will show it in standard range.

Omni Reference: direct with images and videos

Omni Reference is the biggest change for anyone who plans their shots. You can give Kling 4.0 up to 15 references in one generation: photos of your characters, product shots, storyboards, layout sketches, a video whose camera move you want, or a clip of someone performing the action.

From a rough 3D blockout to a finished scene

Kling 4.0 accepts plain grey 3D "white model" animation as a motion reference. Below, the blockout on the left set the action and the camera; reference images of the two characters set the look on the right.

Copy a performance from a depth video

A depth-map video works too. The left clip is only the outline and distance of a moving body; Kling 4.0 kept every move of the kung fu form and dressed it with a character from a reference image, on the Great Wall.

Remake an ad with a new product

Kling calls this creative recreation: keep the pacing, framing and camera of an existing video and swap in your own product or character. The left ad sells a coffee can; with one product image, the right one sells a soft drink in the same shot.

Edit a video you already have

Kling 4.0 can change one thing in an existing clip and leave the rest alone: a character, an expression, the camera, the background or the whole style. You can give it up to 5 clips, 30 seconds in total. Here the corgi in the snow became a voxel toy, and the run, the snow and the camera stayed the same.

Longer stories in one take

With up to 30 seconds per generation, a scene no longer has to stop at the moment it gets interesting. This 28-second example holds one continuous story, from a wide shot of a burning city to a dragon breathing fire over the rooftops.

Up to 10 keyframes

Multi-keyframe input lets you pin the important moments: where a character is, what changes in the scene, how the story turns. Kling 4.0 fills in the motion between them. In this 20-second clip, the keyframes take a glowing neon doodle from a hand to a computer screen, a fridge and across a dark bedroom.

Prompts read like a director

Prompts can now run to 8,000 tokens, and Kling says the model expands short prompts "with a director's eye". Kling's example prompt for this shot asked for "a continuous 7-second shot set in the UK" where two old friends meet at a street corner at dusk, with a handheld camera, warm film look, footsteps, street sound, music and English dialogue.

Text that stays readable

Kling 4.0 writes legible text in many languages, plus emoji and logos, and keeps it steady while the camera moves. That matters for title cards, posters, product labels and social ads.

What to do until Kling 4.0 is on Flixly

• Get your references ready. Character photos, product shots and a storyboard frame or two will be the fastest way to try Omni Reference on day one.
• Try the Kling models that are live now. Kling 3.0 handles text to video and image to video, and Kling O3 Reference builds a clip from your reference images in the reference mode.
• Plan longer scenes in Story Creator. Story Creator already turns an idea into a multi-scene episode, and 30-second takes will give each scene more room.

When Kling 4.0 arrives you will find it in the AI Video Generator, with the credit price shown before you generate. We will confirm then which modes, lengths and resolutions are available on Flixly from launch.