Kling O3
Generates **3 to 15 seconds of video with sound in one pass**, at 1080p in Pro or in full 4K.
Kling O3: Video with Native Sound from Text, Images, and a Reference Clip
Kling O3 is Kuaishou’s newest video model. It generates 3 to 15 seconds of video with sound in one pass, at 1080p in Pro or in full 4K. Give it a text prompt, a first frame, a set of reference images, or a reference video, and tag each file right in the prompt: @Image1 for a subject, @Video1 for the clip whose motion you want.
The reference video is the headline. Attach a clip and O3 follows its motion, framing, and timing while you swap in new subjects, a new setting, or a new style from your images. The result is a new video of the length you choose, with the clip’s sound kept or dropped.
What can you do with it?
- Write a scene and get a finished shot. Prompt it and O3 generates picture and sound together: dialogue, ambient noise, effects.
- Put a real person or product into a new scene. Attach their photo, describe the scene, and O3 keeps the face and look consistent.
- Animate a still image. Your image becomes the first frame and the video grows out of it.
- Control the start and end. Frame mode takes a first and a last frame and generates the motion between them.
- Recast a clip. Attach a video as @Video1 and images of who or what should be in it, and O3 replays the motion with your subjects.
- Restyle a clip. Attach a video and a style image, and O3 keeps the movement while the look changes.
- Cut a multi-shot sequence in one go. Switch Shots to Auto and O3 plans the cuts from your prompt instead of holding one continuous take.
How to use it
- Pick your mode. Omni Mode takes a prompt plus reference images and at most one reference video. Frame Mode takes a first frame, an optional last frame, and a motion prompt.
- Attach your references. Up to 7 images on their own, or one video plus up to 4 images. Images should be at least 300 px on each side. The video should be 3 to 15 seconds, MP4 or MOV, 720p or better.
- Write the prompt. Call out each reference by its tag and say what it contributes: “@Image1 is the dancer, @Video1 is the choreography.” Say what should stay and what should change.
- Set model, duration, and ratio. Standard, Pro, or 4K. 3 to 15 seconds. 16:9, 9:16, 1:1, or Auto to follow your reference video.
What you get
One video, 3 to 15 seconds, as mp4. Audio is generated with the video, or taken from your reference clip when you attach one and keep its sound. In frame mode the output ratio follows your first frame.
Tips
- Name the role of every reference. “@Image1 is the hero, @Video1 is the camera move” beats attaching files and hoping. O3 follows explicit assignments.
- Draft on Standard, finish on Pro or 4K. Same model, same prompt. Iterate cheaply, then rerender the keeper at the resolution you need.
- Use clean, well-lit subject photos. Consistency follows reference quality; a sharp frontal shot holds up better than a blurry crop.
- Match the reference clip’s length. When a video drives the motion, set the duration to the clip’s length so nothing gets stretched or cut short.
- Write dialogue in quotes. With audio on, O3 speaks quoted lines in the prompt and matches the lips to them.
- For a precise ending, switch to Frame Mode. A last frame pins the final composition harder than any prompt can.