Why Kling AI Raised at $3B: How AI Video is Actually Being Used
Early AI video was full of melting faces and weird physics. The new wave of video models keeps characters and camera movements steady, which is why commercial ad agencies and studios are actually paying for them.
What changed with AI video?
A year ago, generative AI video was mostly a party trick on social media. People posted funny clips of Will Smith eating spaghetti, characters with six fingers, and people morphing into furniture.
Today, companies like Kling AI are raising money at multi-billion-dollar valuations, and commercial ad agencies are actively using these tools in client campaigns.
The difference isn't that the models suddenly became magical. It is that they finally fixed two specific problems that made earlier models unusable for commercial work: character consistency and camera control.
The practical changes that make it usable
1. Real camera movement controls With older video generators, you typed a prompt like "slow pan left," and the model would often zoom in, blur the subject, or spin wildly.
Newer models let you pick specific camera movements—like a steady dolly shot, a smooth tilt, or a clean pan—using simple controls. You can set a start frame and an end frame, and the model fills in the motion smoothly between them.
2. Faces and objects stay the same If an actor turns their head in a scene, their clothes, haircut, and facial features used to morph into an entirely different person. Newer architectures keep the subject recognizable across consecutive cuts, which is the bare minimum required for any product advertisement or brand story.
Where the money is actually going: Social ads and variation
Hollywood is not replacing full feature films with a single prompt. Where this technology is actually making money right now is digital advertising.
Shooting a traditional product commercial takes weeks of planning, studio rentals, actor hiring, and expensive color grading. If the client wants to test three different background locations or adjust the lighting, the production crew has to go back on set.
Creative agencies now use models like Kling to generate dozens of localized ad variations in an afternoon: * Swapping the background from a city street to a beach scene. * Testing different product colorways in realistic lighting without manufacturing physical prototypes. * Producing short 5-to-10 second video hooks for TikTok and Instagram ads at a fraction of standard production costs.
What's still hard
AI video still has clear limits: * Audio and speech: Lip-syncing and sound design are still usually handled by separate specialized tools. * Complex physical interactions: If a character needs to tie their shoelaces or unlock a padlock with a small key, the model often struggles with the fine motor physics. * Long stories: Making a full two-minute video that flows seamlessly across ten different shots still requires a skilled human editor sitting in Premiere or DaVinci Resolve.
The tools are not replacing the human director—they are speeding up the tedious parts of getting an idea from a storyboard onto a screen.
Startups Cult Research
Frontier Tools & Labs
