Dublê vs Kling 4.0: which one to use to swap a character in a video
8 min read
Kling 4.0 is a general video model that makes scenes from text and images and, per Kling, edits footage that already exists. Duble AI (written Dublê in Portuguese) does one job: it puts your character into a trend or your own video, with the original sound and no prompt. For a new scene or full control, pick Kling; for the same video with a different character, pick Duble AI.
Duble AI is our product, and we did not run Kling 4.0, so this page gives no verdict on quality. What comes from Kling was read on its pages and in reports, each cited where it appears, on October 20, 2026. What the pages do not say is marked "not stated".
Duble vs Kling 4.0: what is the difference?
Kling 4.0 serves many jobs; Duble AI serves one. The table sums up what each documents.
| Kling 4.0 | Duble AI | |
|---|---|---|
| Kind | General video model, from Kuaishou | An app that does one job: a character inside a video |
| Starting point | Text, images, keyframes, reference videos | A trend from the library or your own video |
| New scene from text | Yes | Never |
| Edits existing video | Yes, per Kling: add, replace or remove elements | It is the only job: change who appears |
| Inputs per generation | Up to 15 references (10 images, 5 videos, 7 subjects) | One video and one character |
| Length | 3 to 30 seconds per generation | Your own video from 3 to 30 seconds |
| Instructions | Prompt of up to 8,000 tokens | None required; you can say who is replaced |
| Sound | Stereo audio and voice references; what happens to an edited video's sound is not stated | The original video's; the app never creates music or a voice |
| Price | Credits per second; the 4.0 rate is not stated on the pages read | 3 credits per second at Standard |
| Character across videos | Reference subjects, up to 7 per generation | The character is made once and reused |
Kling sources: the Kling 4.0 guide of September 30, 2026, and Kling's 4.0 and 3.0 comparison.
What is Kling 4.0?
It is the newest generation of video models from Kling, owned by Kuaishou. According to 36Kr, it officially launched on the night of October 15, 2026, after early access for a limited group that began on September 28.
The Kling guide describes:
- native generations of 3 to 30 seconds, with 4K output and a 21:9 ratio;
- up to 15 references per generation: up to 10 images, up to 5 videos (30 seconds combined) and up to 7 subjects;
- up to 10 keyframes on the timeline and prompts of up to 8,000 tokens;
- reference videos that guide movement, camera and pacing;
- two-channel stereo audio and voice references that define a character's voice.
fal.ai, which sells access to models, also lists text-to-video, image-to-video and first and last frame (September 30, 2026).
Can Kling 4.0 replace the character in a video?
By Kling's description, 4.0 edits footage that already exists. The guide says you can "add, replace, or remove elements" and change backgrounds, using text and visual references. An edit takes up to five input videos, one of them the primary, with 30 seconds combined.
Kling also describes recreating popular formats: take a video's pacing and composition and put your own characters, products or other assets in it. The example swap prompt in the 4.0 and 3.0 comparison is "Replace the video's environment and characters entirely with the content from the image": you describe the swap in text.
Motion Control, Kling's separate tool for moving a character with the movement of a video, appears on the pages read only with versions 2.6 and 3.0. It takes one main subject and does not say whether it keeps the reference's audio. See Kling Motion Control: limits, price and alternatives.
Does Kling 4.0 keep the original sound of a video?
Not stated. The 4.0 guide mentions higher-quality audio, stereo and voice references, and the 4.0 and 3.0 comparison mentions more accurate lip sync, but neither says what happens to the soundtrack of an edited video.
In Duble AI the sound of the result is the original video's. The app asks what plays in the video (someone speaks or sings, or only music and background sound), and either way the sound is the video's own. The app never composes music or creates a voice. If you want new dialogue or music generated with the picture, you need a text-to-video model such as Kling. The difference between the two jobs is in text-to-video or character swap.
How much does Kling 4.0 cost next to Duble AI?
Kling 4.0's credit price is not stated on the Kling pages we read on October 20, 2026, and fal.ai said its own price would go up on the model page at launch. For reference, the Motion Control guide charges 9 credits per second on Kling 3.0 in Standard mode: 90 credits for 10 seconds. On Kling's Standard plan (US$10 for 660 credits at list price, read on October 13, 2026) that comes to about US$1.36 per clip. Kling shows promotional prices, so read the plans page before subscribing.
On Duble AI the charge is by the second and shown before you generate:
- 10 seconds at Standard quality (720p) use 30 credits;
- on the larger monthly plan that video comes to $4.44, or $0.44 per second;
- your own video pays the same per second as a trend;
- if a video fails the credits come back, and credits never expire.
Kling's prices are in dollars. Duble AI's appear in the currency of where you are: reais in Brazil, with Pix or card, and dollars elsewhere. The full breakdown is in how much Duble AI costs.
Which one for which job?
The pair above shows Duble AI's job: the original trend and the same video with a different character, with the same sound. The app warns that the scene can change a little. Kling fits better when there is no starting video.
| I want | Use | Why |
|---|---|---|
| A scene that does not exist yet, described in words | Kling 4.0 | Duble AI never creates a scene from text |
| A trend with my character and the original music | Duble AI | The trend is the source video and the sound stays |
| My own video with another character in my place | Duble AI | Your scene and your movement stay |
| Swap a product or the background of a video, or restyle it | Kling 4.0 | Duble AI only changes who appears |
| Up to 30 seconds with many references and keyframes | Kling 4.0 | Fine control that Duble AI does not offer |
| The same character across many videos, with no prompt | Duble AI | The character is made once and reused |
| New dialogue or music generated with the picture | Kling 4.0 | Duble AI never generates sound |
If your job is getting an AI influencer into trends, see how to make AI trend videos and how to replace the person in your own video. For all the tools that swap a character, there are the AI character swap video tools, and for general models, the AI video models in 2026. To try it, create your character.
Is Kling available outside the US and in other languages?
Kling publishes pages in Portuguese, such as the Motion Control page, and says 4.0 widens its language support, Portuguese among them. The plans page we read is in dollars. Local-currency prices and Pix are not stated. Duble AI opens in the reader's language, charges in reais in Brazil with Pix, and in dollars elsewhere.
What has Kling announced for after 4.0?
Only what Kling itself has published, and only what is on the pages read. Nothing below is a promise from Duble AI.
- 10-bit HDR at 1080p and 4K. Kling's guide, dated September 30, 2026, says "coming soon". Reports at launch disagree on whether it is already live, so check with Kling.
- Video extension, up to 2 minutes. fal.ai, on September 30, 2026, cites Kling's release notes to say an extension would cap extended video at 2 minutes. It is not in the Kling guide we read.
- API for businesses and developers. LetsDataScience, on October 16, 2026, attributes to Kuaishou an API still to come, with no specification or price. We could not open Kuaishou's own release, so treat this as a report.
We found no Kuaishou plan for a version after 4.0.
Frequently asked questions
- Can Kling 4.0 replace the character in my video?
- Kling says 4.0 edits footage that already exists: you can add, replace or remove elements, with up to five input videos and 30 seconds combined. You write the instructions and send the references. Kling's pages we read describe it as an edit made from text and references.
- Does Kling 4.0 keep the original audio of a video?
- The Kling pages we read do not say. They describe two-channel stereo audio, voice references and better lip sync, but not what happens to the soundtrack of an edited video. In Duble AI the sound of the result is always the original video's, and the app never creates music or a voice.
- What is the difference between Duble AI and Kling 4.0?
- Kling 4.0 is a general model: it makes video from text and images and, per Kling, edits existing video. Duble AI does one job: it puts your character into a trend from the library or into your own video of 3 to 30 seconds, with no prompt, and never invents a scene from text.
- Is Kling 4.0 out, and what comes after it?
- According to 36Kr, Kling 4.0 officially launched on the night of October 15, 2026. Kling's September 30 guide marks 10-bit HDR at 1080p and 4K as coming soon. Reports also mention an API for businesses and video extension, with no dates.
- How much does a video cost on Duble AI compared with Kling?
- On Duble AI, 10 seconds at Standard quality use 30 credits, or $4.44 on the larger monthly plan. In the pages we read, Kling has not published a credit price for 4.0. Kling 3.0 Motion Control costs 9 credits per second in Standard mode.
- Can I use Kling and Duble AI together?
- In principle, yes. A video made in Kling is a video file, and if its terms allow it you can upload it to Duble AI as your own video, from 3 to 30 seconds. Upload only videos you have the right to use, as the Terms ask.
