Kling 4.0
Kuaishou's Kling 4.0 turns a prompt, a first frame, or up to 15 reference images and clips into 3 to 30 seconds of video with stereo sound, lip-synced dialogue and multi-shot cuts, at up to 1080p and ultrawide 21:9.
Kling 4.0: Up to 30 Seconds of Video with Sound from Text, Images and Clips
Kling 4.0 is Kuaishou’s newest video model, and you can use it on Mitte right now. Write a scene and get it back as a finished shot with its own sound: dialogue, ambient noise, music, effects. Attach a photo and it becomes the first frame. Attach a clip and the model follows its motion, camera and pacing while you swap in new people, products or settings from your images.
The big change is length. A single run is 3 to 30 seconds, long enough for a multi-shot sequence with cuts the model plans itself. Give it one prompt and it builds the shots, or hand it a shot list and it follows yours. Either way the characters keep their faces and clothes across cuts.
It also holds on to your references better. Up to 10 images and 5 clips go into one run, tagged @Image1, @Video1 and so on, and you tell it what each one is for: this is the dancer, this is the outfit, this is the choreography. Output at 720p or 1080p, in 16:9, 9:16, 1:1 or cinematic 21:9.
What can you do with it?
- Write a scene and get a finished shot. Prompt it and Kling 4.0 generates picture and stereo sound together. Put dialogue in quotes, name the language, and the lips match the words.
- Animate a still image. Your image becomes the first frame and the video grows out of it. Add a last frame to pin the ending.
- Put a real person or product into a new scene. Attach their photo, describe the scene, and the face and the look stay consistent.
- Recast a clip. Attach a video as @Video1 and photos of who should be in it. The motion, framing and timing replay with your subjects.
- Restyle a clip. Attach a video and a style image, and the movement stays while the look changes.
- Cut a multi-shot sequence in one go. Describe a 20-second story and the model plans the shots, the cuts and the camera moves.
- Shoot ultrawide. 21:9 for cinematic framing, straight out of the model.
- Put text on screen. Titles, logos and captions that hold still through camera moves, in English, Chinese, Japanese, Korean, Spanish, Portuguese, German, French and Hindi.
How to use it
- Pick your mode. Omni Mode takes a prompt plus reference images and clips. Frame Mode takes a first frame, an optional last frame and a motion prompt.
- Attach your references. Up to 10 images and 5 clips, 30 seconds of video in total. Images should be at least 300 px on each side. Clips should be 3 to 30 seconds, 720p or better.
- Write the prompt. Call out each reference by its tag and say what it contributes: “@Image1 is the hero, @Video1 is the camera move.” Then direct the shot: subject, action, setting, camera, light, sound.
- Set duration, ratio and resolution. 3 to 30 seconds. 16:9, 9:16, 1:1, 21:9, or Auto to follow your references. 720p to explore, 1080p to finish.
What you get
One video, 3 to 30 seconds, as mp4 with stereo sound. In Frame Mode the output ratio follows your first frame.
Tips
- Give every reference a role. “@Image1 is the singer, @Image2 is the jacket, @Video1 is the dance” beats attaching files and hoping. Tags without jobs get guessed at.
- Direct the shot. “Slow push-in, warm tungsten light, she turns at the window” lands. “Premium cinematic vibe” gives the model nothing to draw.
- Write dialogue in quotes and name the language. Add the speaker’s age, tone and pace if you want a particular voice.
- Match the clip’s length when recasting. A 12-second reference wants a 12-second output so nothing gets stretched or cut short.
- Let it plan the shots. For a story, describe the beats in order and leave the cutting to the model. Pin the structure yourself only when you need exact shot timings.
- Draft at 720p, finish at 1080p. Same prompt, same references. Find the take, then rerender the keeper.
- For a precise ending, use Frame Mode. A last frame pins the final composition harder than any prompt can.
- Keep negatives short. “No subtitles, no background music” is about all you need. Say what you want to see instead of what you don’t.