The text-to-video AI boom has really kicked off in the past few months, with the only downside being that the likes of OpenAI Sora still aren’t available for us to try. If you’re tired of waiting, a new rival called Dream Machine just landed – and you can take it for a spin right now.
Dream Machine is made by Luma AI, which has previously released an app that helps you shoot 3D photos with your iPhone. Well, now it’s turned its attention to generative video, which has a free tier that you can use right now with a Google account – albeit with some caveats.
The essentials
Video resolution: 1360×752
Clip lengths: Five seconds
Training data: unknown
Cost: free tier gives you 30 generations per month, paid tiers start from $29.99 / month
The main one is that Dream Machine seems to be slightly overwhelmed at the time of writing. There’s currently a banner on the site stating that “generations take 120 seconds” and that “due to high demand, requests will be queued”. Our text prompt took over 20 minutes to be processed, but the results (below) are pretty impressive.
Dream Machine’s outputs are more limited in length and resolution compared to the likes of OpenAI’s Sora and Kling AI, but it’s a good taster of how these services will work. The clips it produces are five seconds long and in 1360×752 resolution. You just type a prompt into its search bar and wait for it to appear in your account, after which you can download a watermarked version.
While there was a lengthy wait for the results (which should hopefully improve once initial demand has dropped), our prompt of ‘a close-up of a dog in sunglasses driving a car through Las Vegas at night’ produced a clip that was very close to what we envisaged.
Dream Machine’s free plan is capped at 30 generations a month, but if you need more there are Standard (120 generations, $29.99 a month, about £24, AU$45), Pro (400 generations, $99.99 a month, about £80, AU$150) and Premier (2,000 generations, $499.99 a month, about £390, AU$750).
A taste of AI videos to come
Like most generative AI video tools, questions remain about exactly what data Luma AI’s was trained on – which means that its potential outside of personal use or improving your GIF game could be limited. It also isn’t the first free text-to-video tool we’ve seen, with Runway’s Gen 2 model coming out of beta last year.
The Dream Machine website also states that the tool does have technical limitations when it comes to handling text and motion, so there’s plenty of trial-and-error involved. But as a taster of the more advanced (and no doubt more expensive) AI video generators to come, it’s certainly a fun tool to test drive.
That’s particularly the case, given that other alternatives like Google Veo currently have lengthy waitlists. Meanwhile, more powerful models like OpenAI’s Sora (which can generate videos that are 60-seconds long) won’t be available until later this year, while Kling AI is currently China-only.
This will certainly change as text-to-video generation becomes mainstream, but until then, Dream Machine is a good place to practice (if you don’t mind waiting a while for the results).
