AIREITER
What Kling 4.0 generates
Kling 4.0 is a video model from Kuaishou Kling. One model covers text-to-video, keyframes, image, video and audio references, video editing, and video extension, and every clip comes with a stereo audio track.
Four modes, one model
Text / Reference, Frames, Edit Video, and Extend Video. Pick the mode by what you already have.
3 to 30 seconds
Any whole-second length from 3 to 30, default 5. In Extend mode the length is the new part only.
720p or 1080p with sound
Sound is generated with the picture on every clip. There is no 4K and no silent option.
16:9, 9:16, 1:1, 21:9
Set the ratio in Text / Reference mode, or leave it on auto. Frames follow the first frame; Edit and Extend keep the source ratio.
Kling 4.0 test results
Both clips are from our own tests on November 4, 2026. Size, length, and sound are measured from the delivered files.
Text to video: paper boat on a neon street
Request: Text only, 16:9, 5 seconds, 1080p.
Delivered file: H.264 MP4, 1920x1080, 5.04 seconds, 24 fps, stereo AAC audio.
Generation time: about 4.5 minutes.
Extend Video: five more seconds of the same clip
Request: Extend Video, 5 new seconds.
Delivered file: 960x540, 10.0 seconds: the source first, then the new part.
What held up: no visible cut at the 5-second mark, and the street, signs, and reflections stay the same. In the new seconds a passer-by walks in and lifts the boat from the water.
Sound: the silent source stays silent, and the new seconds come with generated sound.
Which Kling 4.0 mode to use
Each mode takes different inputs.
Text / Reference
Prompt only for text-to-video. Or add up to 10 images, 5 videos (30 seconds in total), and 3 audio clips of 5 to 30 seconds, 15 items at most, and name them @Image1, @Video1, @Audio1 in the prompt.
Frames
A first frame is required. Add up to 8 middle frames in playback order and an optional last frame: 10 frames at most, and no more than the video length plus 1.
Edit Video
One source clip of 3 to 30 seconds plus up to 10 reference images. Say what to change; the source is @Video1. Keep the original sound or regenerate it.
Extend Video
One source clip of up to 30 seconds, plus an optional last frame to end on. You get back the source and the new part as one file.
What to make with Kling 4.0
A longer shot in one go
Up to 30 seconds per generation, so a product walk-around or a short scene does not need stitching.
Storyboard to motion
Lay out a first frame, middle frames, and a last frame, and let the model fill the motion between them.
Keep a character or product consistent
Give character, outfit, product, and location images, plus a reference video for motion or a reference audio clip for voice.
Rework footage you already have
Change the style, outfit, or background of an existing clip with Edit, or carry it on with Extend, as in the paper boat test.
How to generate with Kling 4.0
The same steps work in the playground and through the API.
Choose a mode
Text / Reference, Frames, Edit Video, or Extend Video.
Add your inputs
References for Text / Reference, a first frame for Frames, one source clip for Edit and Extend.
Describe the video
Say what moves, then camera, light, and sound. For Extend, describe what happens next.
Set length and resolution
3 to 30 seconds at 720p or 1080p. In Extend, the length is the number of new seconds.
Call Kling 4.0 through the AIReiter API
One asynchronous contract for all four modes.
Model ID
Send kling_v4.
Submit
POST /api/openapi/submit with model, a unique out_task_id, and params. Set type to reference, frame, edit, or extend.
Inputs by mode
reference: prompt with optional image_url, video_url, audio_url. frame: image_url (first frame), keyframe_url, end_image_url. edit: video_url, image_url, audio_setting. extend: video_url, end_image_url.
Query
POST /api/openapi/query with the same out_task_id until the status is completed or failed. Completed responses include the video URL.
Related AI video models
Kling v3 Omni
3 to 15 second clips from text, images, first/last frames, or a reference video, up to 4K.
Kling 3.0
3 to 15 second clips with native audio and first/last frame control, up to 4K.
Kling 3.0 Turbo
Fast 3 to 15 second clips from a prompt or a reference image, at 720p or 1080p.
Wan 3.0
2 to 30 second clips from text, first/last frames, mixed references, or a document.
Kling 4.0 FAQ
Common questions about modes, length, references, sound, and the API.
/ 01What is Kling 4.0?
A video model from Kuaishou Kling that makes 3 to 30 second clips with stereo audio at 720p or 1080p. It works from text, keyframes, or image, video and audio references, and it can edit or extend a clip you upload.
/ 02How long can a Kling 4.0 video be?
3 to 30 seconds in whole seconds, default 5. In Extend mode the length counts the new part only.
/ 03What does Extend Video return?
One file with the source followed by the new part. In our test a 5-second source extended by 5 seconds came back as a 10.0-second video.
/ 04Does Extend keep the original sound?
Yes. The source keeps its own sound and the new part gets generated sound. In our test the silent source stayed silent.
/ 05Can I set first, middle, and last frames?
Yes. Frames mode takes a required first frame, up to 8 middle frames, and an optional last frame.
/ 06How many references can I add?
Up to 10 images, 5 videos (30 seconds in total), and 3 audio clips, 15 items in all.
/ 07Can I turn the sound off?
No. Every clip has a stereo audio track. In Edit mode you can keep the original sound or regenerate it.
/ 08How is Kling 4.0 different from Kling v3 Omni?
Kling 4.0 goes up to 30 seconds and adds middle frames, Edit Video, and Extend Video, at 720p or 1080p. Kling v3 Omni stops at 15 seconds but offers 4K.
/ 09Which API model ID should I use?
kling_v4, with POST /api/openapi/submit, then poll POST /api/openapi/query with the same out_task_id.