In recent months, there has been a surge in AI models that generate videos based on text requests. However, most of these advanced tools, like OpenAI Sora, are still inaccessible to the general public. For those tired of waiting, there’s a new competing service called Dream Machine, which has just launched and is ready for use.
Dream Machine is created by Luma AI, previously known for their 3D photo capturing app on iPhone. They have now shifted focus to generative video and offer a free tier accessible to anyone with a Google account — albeit with some caveats.
The main caveat is that Dream Machine currently seems somewhat overloaded. The website displays a banner warning that “generation takes 120 seconds” and “due to high demand, requests will be queued.” According to Techradar, generation times can sometimes exceed 20 minutes.
Videos generated by Dream Machine are more limited in duration and resolution compared to services like OpenAI Sora and Kling AI, but they provide a good preview of how such services will function. The created clips are five seconds long and have a resolution of 1360×752 pixels. Simply enter a query into the search bar, and after processing, you can immediately download the clip with a watermark.
Researchers from Techradar asked the model to generate “a close-up of a dog wearing sunglasses, driving through nighttime Las Vegas.” The result was quite impressive:
The free tier of Dream Machine is limited to 30 generations per month, but if you need more, there are plans available: Standard (120 generations, $29.99 per month), Pro (400 generations, $99.99 per month), and Premier (2000 generations, $499.99 per month).
As with most AI video generation tools, it remains unclear on what data the model used in Dream Machine by Luma AI was trained. Dream Machine is also not the first free tool for generating videos from text descriptions. Last year, after a lengthy beta testing phase, Runway released a similar service, Gen 2.
The company’s website also indicates technical limitations in text and motion processing, so much refinement is still needed.
However, for now, this tool represents one of the few opportunities to try out this technology. Other alternatives like Google Veo currently have long waitlists. More powerful models such as OpenAI Sora (capable of generating 60-second videos) won’t likely appear until later this year, and Kling AI is currently only available in China.
Certainly, the landscape will change when text-to-video generation becomes mainstream, but until then, Dream Machine is an excellent way to practice (if you don’t mind waiting a bit for results).