Game of Catch-Up Over? Kling 3.0 Throws Down the Gauntlet to Sora and Veo with a Unified Model and 15-Second Scenes

The Kling AI team has unveiled a major platform update—the unified VIDEO 3.0 model. Unlike previous versions that split tasks, the new architecture combines text-to-video, image-to-video, reference-based generation, and editing within a single "engine." The main focus is on increasing duration, element consistency, and narrative control without the need for complex editing.

The model generates videos up to 15 seconds long with flexible duration settings, which, according to the developers, allows for fitting complete story-driven scenes into a single clip, not just short snippets. The new Multi-Shot function analyzes a scene description and automatically constructs it from multiple shots: changing angles, framing (e.g., wide shot, dialogue shot, close-up), and following cinematographic patterns. This is meant to free creators from manually stitching scenes together.

The new model also allows "locking in" the appearance of a character or object using one or several images or even a short video (3–8 seconds). After that, the model preserves their features throughout the entire clip, despite camera movement or action changes. Moreover, users can additionally upload an audio track with speech to give a character a specific voice. Overall, audio accompaniment now supports assigning lines to specific characters in scenes with multiple actors. Added support for multiple languages (Chinese, English, Japanese, Korean, Spanish) and dialects, as well as the ability to mix languages in a single dialogue with synchronized lip movements.

As Kling AI notes, the update is built on a completely redesigned multimodal architecture where different data types (text, image, sound, video) are processed in a single pipeline. This allows for more accurate interpretation of complex prompts and ensures output consistency. Early access to VIDEO 3.0 and VIDEO 3.0 Omni is now open to Ultra subscription users. The release for other plans is scheduled for the near future.

Kling AI is clearly targeting creators for whom not only visual quality is important, but also the integrity of the story—from script to sound. Increasing clip length to 15 seconds and built-in control tools put the service in direct competition with solutions like Google Veo, which also bets on long and coherent scenes.

What do you think: how relevant are such systems for the average user if the main goal is to quickly create simple clips for social media? Or is this still a tool for more professional content production? Share your thoughts in the comments.