Kling 4.0

Kling 4.0 is Kuaishou's newest Kling video model. In Masonry it generates video from text, from a start image with an optional end frame, or from up to 10 reference images, and it edits videos you upload according to a text instruction. Clips run 3 to 30 seconds at 720p or 1080p with native stereo audio.

Yes. The video editing endpoint takes a source clip of 3 to 30 seconds and a written instruction, plus optional reference images. Describe the change and what should stay the same, and refer to the source as @Video1 in the prompt. It is the model behind Masonry's video editing in the MCP edit_video tool and the Masonry CLI.

Any whole-second length from 3 to 30 seconds for text-to-video, image-to-video, and reference-to-video. With a multi-shot prompt, the clip length is the sum of the shot durations. An edited video keeps the length of the source clip unless you set a different duration.

Kling 4.0 outputs 720p or 1080p in every mode. Draft at 720p to explore pacing and composition, then render the version you keep at 1080p.

Text-to-video supports 16:9, 9:16, 1:1, and 21:9. Reference-to-video supports the same four plus auto, which adapts the frame to your references. Image-to-video follows the start image, so crop the source to the placement before you generate.

Yes. fal says Kling 4.0 generates dialogue, singing, sound effects, and ambience as two-channel stereo in sync with the picture, with speech in languages including Chinese, English, Japanese, Korean, and Spanish. The endpoints have no audio switch, so audio is always on. Describe the sound you want, or mute the track in your editor if you plan to add a voiceover.

Reference-to-video accepts up to 10 reference images. Address each one in the prompt as @Image1, @Image2, and so on, and give each referenced subject a clear role so the model knows which image is the person, which is the product, and which is the setting.

Yes. Image-to-video takes a required start image and an optional end image, and the model moves from one to the other. Two approved frames fix where the clip begins and lands, but the motion between them is generated, so review the whole clip, not just the endpoints.

A multi-shot prompt splits one generation into shots, each with its own prompt of up to 512 characters and its own duration. The video runs for the sum of the shot durations, which lets you direct a short sequence, such as an establishing shot, a detail, and an end frame, in a single render.

In Masonry, Kling 4.0 extends the clip ceiling from 15 to 30 seconds, generates stereo audio with every clip, takes up to 10 reference images, and adds editing of an uploaded video. Kling O3 Standard is still available in Masonry for 3 to 15 second clips with optional audio.

Yes, with the usual review discipline. Use an approved product photo as the start image or as a reference, keep the motion restrained, and watch the full clip, then inspect evenly spaced frames against the exact SKU. Check the silhouette, label text, logo, color, material, part count, and quantity, and reject any clip that alters the sold item or shows a surface your sources never did. Generated in-between frames can illustrate a transition but cannot prove how a mechanism works. Keep prices, claims, offers, and disclosures out of the generated pixels and speech, and add them as overlays you can approve and update.

Masonry retired both in November 2026, and their pages now redirect here. Kling 4.0 covers what each did, start and end frames with multi-shot direction like O3 Pro, and scenes built from references addressed with @ tags like O1 Reference (Character Lock), with longer clips, more references, and stereo audio. Kling O3 Standard is still available.

Masonry bills Kling 4.0 in credits by resolution and clip length, so a 1080p or longer clip costs more than a short 720p draft. The editor shows the exact credit cost before you generate.

Yes. Install the Masonry CLI with npx @masonryai/cli and sign in with masonry login. Run masonry models params kling-v4 to see the inputs the route accepts today, then generate or edit from the terminal. Video editing in the CLI uses Kling 4.0.

fal marks all four Kling 4.0 endpoints for commercial use. Review the current fal and Kuaishou terms for your project, and get consent before generating or editing a real person's likeness or voice.

Open a Masonry canvas, pick Kling 4.0 from the model selector, and describe the video you need: a product shot, an ad creative, a social post. Masonry generates it, then you refine, edit, and combine Kling 4.0 with other models in one workspace.

New Masonry accounts get 50 free credits to try GPT Image 2.5 Flare. Kling 4.0 is included in every paid plan, and the editor shows the exact credit cost before you generate.

Write the prompt like a short shooting script. Name the subject, the action, the camera move, the setting, and what should be heard, and put each sound beside the moment that causes it. For reference-to-video, address every input by its tag, for example "@Image1 holds @Image2 up to the camera", and give each referenced subject one clear action. For multi-shot work, keep each shot to one beat and budget the durations so they add up to the length you want. For an edit, describe only the change and what must stay the same, and refer to the source clip as @Video1. Because audio is always generated, say "no dialogue" or "ambient sound only" when you plan to add a voiceover later. See the prompt gallery on this page for real Kling 4.0 prompts you can copy and adapt.

Kling 4.0 is built by Kuaishou. Inside Masonry it runs alongside 50+ image and video models, so your team can pick the right one for each brief without switching tools.

Yes, the prompt gallery on this page shows real videos teams have generated with Kling 4.0 in Masonry, each paired with the exact prompt you can copy and adapt for your own brand.