Kling 2.6
Kling AI's first model with native audio, launched December 3, 2025. On Kling4 it runs only as Motion Control: you supply a character picture and a clip of someone moving, and Kling 2.6 makes your character move the same way.
- Motion Control only
- Driving clip 3–30 s
- 2 credits/s at 720p
- 3.2 credits/s at 1080p
Where Kling 2.6 sits in the Kling lineup
Kling AI released Kling 2.6 on December 3, 2025 and called it its first native-audio model: one pass produced the picture together with voices, sound effects and room ambience. It came in two forms, text to audio-visual and image to audio-visual, with 5 or 10 second clips and speech in Chinese or English. Lines written in other languages were spoken in English.
Two additions followed within two weeks: Voice Control on December 16, then Motion Control on December 17. In the lineup, Kling 2.6 sits between Kling 2.5 Turbo (September 2025) and Kling 3.0, rolled out on January 31, 2026; Kling's 3.0 guide describes VIDEO 2.6 as upgraded to VIDEO 3.0.
Kling4 can only call the Motion Control part of that release, so that is the Kling 2.6 you can run here. Kling 2.6 text-to-video and image-to-video are not available on Kling4, and the dialogue demos further down are Kling AI's, not runs made on this site.
Kling 2.6 Motion Control on Kling4
- Character picture
- 1 image
- JPG or PNG, up to 10 MB, at least 300 px, between 1:2.5 and 2.5:1.
- Driving clip
- 3–30 s
- MP4 or MOV up to 100 MB, filmed as one take without cuts.
- Character Orientation
- Image or Video
- Image keeps the character facing as in your picture and caps the clip at 10 s. Video turns with the performer and allows the full 30 s.
- Resolution
- 720p · 1080p
- Credits per second
- 2 · 3.2
- Base rates, counted on the driving clip's length: 8 s at 720p is 16 credits, 30 s at 1080p is 96. The price beside Create follows your plan.
- Sound
- From your clip
- Kling4 keeps the driving clip's audio. Motion Control writes no new dialogue or effects.
Motion Control runs need a signed-in account and credits. Searching for Kling 2.6 free? It isn't part of the no-login trial; the one free clip per day on Kling4 is a 5-second Kling 4.0 run.
A Kling 2.6 Motion Control run, input by input
Illustrated student takes over a vlogger’s gestures
An illustrated student on a tree-lined street, driven by a clip of a woman talking to camera. The pointing hands and mouth shapes come from the clip; the braid, cardigan and street come from the picture.
Working with Kling 2.6 Motion Control
- Test cheap, then render. At 2 credits a second, a 4-second slice of your clip costs 8 credits at 720p. Check that the character holds up on a short slice before paying for the full take at 1080p.
- Switch orientation for long clips. Kling's own app starts 2.6 Motion Control on Video orientation; Kling4 starts on Image, which stops at 10 seconds. Change it to Video before uploading anything longer.
- Put the voice in the clip. Lip sync is on the list of things 2.6 transfers, and the output plays the driving clip’s sound. For a talking character, film the lines into the clip itself.
- Drawn characters work. Kling's own 2.6 demo drives an illustrated character with live-action footage. In that demo the street behind her comes from the picture too, so choose a picture whose setting you want to keep.
- Leave the prompt empty if you like. The two files are enough for a run. Add a note of up to 2,500 characters only when something should change, such as the lighting or a prop.
Kling 2.6 native-audio demos (not on Kling4)
Burnt-egg breakfast, two voices
Prompt A kitchen in the morning, sunlight streaming through the window onto the countertop, with a frying pan sizzling. [Boyfriend] places a blackened fried egg on the table, raising an eyebrow proudly. [Boyfriend, cheerful voice]: "Try my breakfast made with love!" During this, [Girlfriend] remains silent. Immediately, [Girlfriend] leans in, takes a light sniff, and raises an eyebrow. [Girlfriend, teasing voice]: "The love is definitely felt, it's just a bit burnt." During this, [Boyfriend] remains silent. Then, the two make eye contact, and together, both smile and say: "It's just a bit burnt." The camera cuts from a close-up of the fried egg to [Boyfriend and Girlfriend] sharing a smile.
Rap verse by a graffiti wall
Prompt Visual: Brooklyn, New York – in front of a graffiti-covered wall, the street vibe is intense, with breakdancers freestyling nearby.Subject: An African-American rapper wearing a gold chain and an oversized hoodie, grooving to the beat while facing the camera.Audio: [African-American rapper, energetic male voice] Rapping over a drum beat: "Yeah, from the bottom to the top, I’m shining bright like a star. Brooklyn streets raised me tough, fought through the dark. Gold chain swingin’, flow hits hard, grindin’ daily, never bored. Now I’m livin’ in the light, this is my life, raw and hardcore!"Background: Layered with deep bass and turntable scratches. Camera cuts rapidly between close-ups of his facial expressions, hand gestures, and the breakdancers.
Lipstick pitch on a livestream
Prompt In a beauty live-streaming room, warm yellow lighting illuminates the table, with lipstick samples displayed on either side.[Caucasian beauty influencer] raises a matte dusty rose lipstick. [Caucasian beauty influencer, sweet and fresh voice] says: "Perfect for yellow undertones! Brightens the complexion without drying, and the finish looks beautifully soft all day." Background: Soft beauty BGM playing.
When Kling 2.6 is the right motion model
Pick Kling 2.6 when the job is a straight copy of one performance onto one character and the budget is tight: at 2 credits a second of driving clip, it is the cheapest way to move a character on Kling4.
Kling 3.0 Motion Control accepts the same two files at 3 credits a second. Kling's March 4, 2026 note names facial element binding as what 3.0 adds over 2.6; Kling4 does not expose that feature, so here the two take identical inputs and differ in model version and rate.
Kling 4.0 with a reference clip is a different tool. It writes a new scene, with new sound, that borrows timing or camera movement from your footage rather than copying one performer's moves; start that from the Kling 4.0 generator. All three routes are compared side by side on Kling Motion Control.
Kling 2.6 questions
Why does Kling4 offer Kling 2.6 only as Motion Control?
Kling4 generates through an API that exposes Kling 2.6 Motion Control but not Kling 2.6 text-to-video or image-to-video, so those two modes cannot run here. For prompt-only clips with spoken dialogue, use Kling 3.0 or Kling 4.0 on Kling4 instead; both generate sound with the picture and can start from a still image.
Kling 2.6 is a native-audio model, so does Motion Control add new sound?
No. Native audio belongs to the text and image modes of Kling 2.6. In Motion Control, Kling4 keeps the soundtrack of your driving clip and plays it under the result, so any voice, music or footsteps you hear come from your own recording. To change the sound, edit or replace the audio in the clip before you upload it.