Kling 3.0 on Morphic: features, multi-shot video, and native audio

How to use Kling 3.0 on Morphic

1. Select video mode

From the prompt bar, select Video mode. This switches the interface to video generation, where you can configure settings like resolution, duration, and whether to include native audio.

2. Select Kling 3.0 as your model

Open the model dropdown and choose Kling 3.0 from the list of available video models. Morphic offers multiple video models, so you can compare outputs across different generators without switching platforms.

3. Add your prompt

Describe the scene you want. Include details about the subject, environment, camera movement, lighting, and any dialogue. Think like a director, not a photographer: describe what happens over time, not just a static frame. If you want multiple shots, turn on the multi-shot toggle or label each shot in your prompt.

4. Generate

Run the prompt. Kling 3.0 produces video clips from 3 to 15 seconds in length, with native audio included when enabled. Review the output, adjust your prompt if needed, and regenerate until you have the clip you want.

What is Kling 3.0?

Kling 3.0 is Kuaishou's latest AI video generation model, released in February 2026. It builds on the Kling Video 2.6 and Kling O1 models by merging them into a unified multimodal architecture that handles video, audio, and text in a single generation pass.

Where earlier AI video models produced isolated single-shot clips with no audio, Kling 3.0 generates multi-shot sequences with synchronized dialogue and native audio output. The model understands cinematic language (tracking shots, close-ups, shot-reverse-shot) and can plan scene transitions on its own when you describe a narrative in your prompt.

Kling 3.0 is available on Morphic as part of the platform's multi-model video generation suite, which means you can use it alongside Morphic's image, music, and audio tools in the same workspace.

Kling 3.0 features and capabilities

Multi-shot storyboard generation with two control modes

This is the feature that separates Kling 3.0 from every other AI video model available right now. It generates up to six camera cuts in a single generation, and it gives you two ways to control them:

• Auto multi-shot: turn on the multi-shot toggle and the model plans the shot transitions itself based on your prompt. It reads your scene description and decides where to cut, what angle to use, and how to pace the sequence.
• Custom multi-shot: you define each shot manually. Set the number of shots, the duration of each, the camera angle, and what happens in the frame. The model follows your storyboard exactly.

Auto mode works well when you want fast results from a narrative prompt. Custom mode is better when you need precise control, for example when building a product ad with specific shot-by-shot pacing.

Element references with voice binding

Most AI video models let you upload a reference image to anchor a character's look. Kling 3.0 goes further. You can upload a short video clip as a reference, and the model extracts both the character's visual appearance and their natural voice tone. That voice gets bound to the character element, so every time that character speaks in your video, the voice stays consistent without you needing to specify it again in the prompt.

You can also create elements from 2-4 reference images and separately assign a voice tone by uploading audio or selecting from available voices. This is especially useful for recurring characters across multiple video generations.

Dialect, accent, and multilingual code-switching

Native audio in Kling 3.0 supports five languages: English, Chinese, Japanese, Korean, and Spanish. But it goes deeper than basic language support. The model can replicate specific dialects and accents, including Cantonese, Northeastern Chinese, Sichuanese, and Beijing dialect for Chinese, and American, British, and Indian accents for English.

It also handles code-switching, meaning characters can switch between languages mid-conversation within the same video. A bilingual business meeting, a tourist asking for directions in broken Spanish, or a family scene mixing dialects all generate with natural lip movements and coherent facial expressions.

Native 4K video output

The model supports up to 4K resolution natively, not upscaled from a lower resolution. This means textures, skin detail, and fine elements like fabric weave and hair strands carry authentic detail rather than the soft, smoothed-over look that upscaling produces. Lower resolutions (1080p and 720p) are also available, and the model supports 16:9, 9:16, and 1:1 aspect ratios.

Text and logo preservation during camera motion

Kling 3.0 can read text from uploaded images, like signs, product labels, or logos, and keep that text legible throughout the video even as the camera moves. It can also generate new text content within the video itself. For commercial work where brand text needs to stay sharp during a product orbit or tracking shot, this removes the need for text overlays in post-production.

Character consistency across multi-shot sequences

Upload reference images or a short reference video, and the model locks a character's appearance through the entire clip. Faces, outfits, proportions, and distinguishing details hold steady through camera movements like zooms, pans, and tilts. The model supports three or more distinct characters in the same scene without blending their features, which matters for dialogue scenes and any video with multiple people.

Flexible duration from 3 to 15 seconds

Generate anywhere from 3 to 15 seconds of continuous video in a single pass. The extra length is not just about longer clips. It gives the model room to develop more complex action, build scene transitions, and let a narrative arc play out rather than cutting short at the five-second mark.