In a long-expected announcement, Luma AI has finally unveiled what’s coming next for the Palo Alto team: a state-of-the-art video model called Dream Machine.
Luma’s CEO, Amit Jain, spoke at Tech Week last week and gave some hints about the upcoming release, but no one expected it to be released just eight days later.
This release doesn’t appear to be related to Radiance Fields or their capture app, but it does seem to answer the question of what Luma has been working on. This is the first major release since the announcement of their generative research project, Genie.
Dream Machine can function from two types of inputs: either from text or from an input image. It can also create temporally consistent video. It appears to be quite strong at creating realistic human faces and motions.
Access to the new platform is now available and is free to use. There has been no word yet on what the licensing terms will be to use Dream Machine.
Earlier this year, Luma raised $43M from a16z with the goal of creating multimodal AI. Over the last several months, Luma has gone on a hiring spree, adding team members deeply experienced in generative AI.
The release comes at an interesting time, with the seemingly competing company Pika just completing an $80M fundraising campaign and the release of the Chinese video model, Kling. There is also, of course, Sora from OpenAI and Veo from Google. Now that Luma has formally entered the video model fray, this appears to be a new era and extension for Luma AI.
To begin trying Dream Machine, head over to the Luma AI website.
