AI video generators are still in their relative infancy, and the most powerful-looking models so far — OpenAI's Sora and Kuaishou's Kling — remain available only to a limited group of testers. But now a San Francisco-based startup has changed that with the unveiling of its Luma AI Dream Machine.
The browser-based AI text video generator is free to try, with no waiting list. And considering where AI videos were just a year ago, it looks like it can produce some pretty shocking footage. While the output remains strange and creepy, the image quality is surprisingly clean and the video generation is fast. It even looks like it might replace our current pick in our guide to the best AI image generators.
Luma AI's Dream Machine works a lot like the best AI image generation tools: you enter a text prompt, and it generates a five-second video in about two minutes. I've been impressed with the initial results; beta testers already using the program say it follows the prompts accurately and generates relatively consistent movements. Characters are consistent and seem more capable of expressing emotion than those seen in many of the AI-generated videos.
The announcement trailer for my new feature film project “Gala”. Created using the just-announced @LumaLabsAI video AI model #LumaDreamMachine. This truly is a game-changing moment for AI video. Thanks to Luma Labs for letting me share the release! pic.twitter.com/ynUNpFytHMJune 12, 2024
Luma AI said it plans to release APIs and plugins for creative software. A public release could also encourage Open AI to accelerate the public launch of Sora, since getting in early puts Luma in a better position to build a community around the platform.
(Image courtesy of Luma AI)
The reality is, despite Showrunner's launch into becoming the “Netflix of AI,” AI-generated video remains largely an experimental curiosity: That is, while it's possible to tell a story using only silent, panning shots of moving, yet somewhat hallucinatory, things that are rarely seen once the curiosity value of the spooky uncanny valley wears off.
The Dream Machine still has its limitations (Image courtesy of Luma AI)
Luma AI is aware of the current limitations of its generator: scenes tear and morph into something else, movement appears unnatural, subjects slide through the scene like ghosts, text appears garbled, polar bears sometimes have two heads, but as mentioned before, it's still early days.
For more AI news this week, check out the release of Stable Diffusion 3 Medium.
Receive daily design news, reviews, how-tos and more from our editors.
