What Kling 4.0 does
Kling 4.0 picks its mode from what you give it. A prompt alone: text to video. A start image, and an end image if needed: the video starts from that image and keeps its framing. References, up to 10 images and 5 videos cited in the prompt: the model composes with them. A fourth mode, Edit, reworks an existing video.
3 to 30 s
per clip, set before you run
1080p
by default, or 720p
Sound
always included, nothing to switch on
The four steps
Step 1
Pick the mode that matches what you have
In Generate, pick "Kling 4.0", then the Frames, Reference or Text type, or Edit to rework an existing video. On the canvas, add the node and connect your inputs: the mode follows what is connected; to rework a video there, choose the "Kling 4.0 · Video Edit" model.
Step 2
Add reference images and videos
In Reference mode, add up to 10 images and 5 videos, then cite them in the prompt as @Image1, @Image2… and @Video1… Each video must last at least 3 seconds, and 30 seconds at most in total. For a video imported into or made in Beemm Vision, going over is refused before anything is charged.
Step 3
Write the prompt
The prompt is required in every mode. Describe the scene, the motion and the sound you expect: the video always comes with a soundtrack, there is no audio option to switch on.
Step 4
Set duration, resolution and format, then run
Duration from 3 to 30 seconds, 720p or 1080p, format 16:9, 9:16, 1:1 or 21:9. Auto, the default in Generate, lets the model frame in Reference mode and gives 16:9 in Text mode; in Frames, the framing follows the start image. The cost shows before you run. Allow several minutes for a 30-second clip in 1080p.
The settings at a glance
| Setting | Values | Tip |
|---|---|---|
| Mode | Text, Frames, Reference, Edit | Edit: the "Edit" type in Generate, the "Kling 4.0 · Video Edit" model on the canvas |
| Duration | 3 to 30 s | Set it yourself: Generate starts at 3 s; in Edit, it follows the source video |
| Resolution | 720p, 1080p | 1080p by default, the most expensive tier |
| Format | 16:9, 9:16, 1:1, 21:9 | Auto by default: 16:9 in Text; in Frames and Edit, the framing follows the input |
| References | 10 images, 5 videos | Cited as @Image1… and @Video1…; 30 s of video at most |
Use cases
A vertical ad with sound
The product as @Image1, the set as @Image2, 9:16, 15 to 30 seconds: the soundtrack is already there.
A transition between two images
Frames mode with a start and an end image, generated earlier in the same graph.
Draw on the motion of a clip
A clip as @Video1, a character as @Image1, and a prompt saying what to take from it. The input video seconds are billed too.
Rework an existing video
In Edit mode: "replace the jacket in @Video1 with the one in @Image1". The output lasts as long as the source.
Frequently asked questions
- Does Kling 4.0 generate sound?
- Yes: every generated video (Text, Frames, Reference) comes with a soundtrack. There is no option to turn it off; describe the sound you expect in the prompt.
- What is the maximum length with Kling 4.0?
- 30 seconds per clip, at least 3. In Edit mode, the output lasts as long as the source video, 3 to 30 seconds.
- How do I cite references in the prompt?
- As @Image1, @Image2… for images and @Video1… for videos, in the order of the image strip in Generate: up to 10 images and 5 videos, 30 seconds of video at most in total (through the MCP server: 8 images and 4 videos).
- How is Kling 4.0 billed?
- Per second, by resolution: 1080p costs more than 720p. As soon as a video goes in (reference videos or Edit), the per-second rate drops, but you pay for the larger of input plus output seconds, or twice the output seconds. The cost shows before you run.
- Where can I use Kling 4.0 on Beemm Vision?
- In Generate ("Kling 4.0", with the Frames, Reference, Text and Edit types), on the canvas as a node, and from an agent through the MCP server. No Kling subscription needed.
Try Kling 4.0
Open the studio, pick Kling 4.0 and run a first 5-second clip in 720p.