I was intrigued by OpenAI's Sora text-to-video AI model but never got around to trying it out for myself. That's not the case with Luma's new Dream Machine video generator, which debuted this week and is now available for anyone to try.
As explained in the Luma blog, the Dream Machine is built on a “scalable, efficient, and multimodal transformer architecture” and is trained on videos. You can provide Luma's Dream Machine with text prompts and images to generate videos. Luma says these videos contain “physically accurate, coherent, and action-packed scenes.”
I've tried to create action-packed videos myself using simple prompts, without much success, but others have been showing off just how capable Luma's Dream Machine really is. X (formerly Twitter) has a ton of amazing examples, but here are a few of my favorites:
With any of these AI video makers, detailed prompts are key and you might need to repeat them a few times to get the results you want. That said, the fact that this is free for everyone is notable and will no doubt encourage more people to give it a try.
Beyond the impressive video samples above, one of the coolest features of this model is the ability to turn photos into videos, which are similar to Apple's Live Photos, but instead of seeing what actually happened, you can make the subject of the photo do virtually anything.
If you want to try out Dream Machine, go to the Luma website, log in with your Google account, and enter your prompt in the text box. Once you're done, tap the arrow on the right and your request will be queued. On average, it took about 3-4 minutes to generate my video. However, keep in mind that you can only generate 30 videos per month for free. If you want to create more videos after that, you'll need to pay.
