Our first video model, Dream Machine, was released in 2024 after a year and a half of building infrastructure and training systems, and it gained significant attention, making Luma a household name.

Despite increasing GPU resources, the intelligence improvements from the current generation of LLMs are not meeting expectations, with reports of failed training runs and diminishing returns emerging from major labs. Additionally, the lack of high-quality data to train new models has become a major bottleneck for AI development.

We are not just a video model company; our goal is to achieve multimodal general intelligence, enabling users to create entire worlds.

I left Apple to start Luma AI in 2022 after realizing that convincing a large organization to adopt new technology was unlikely.

AI video models are finally starting to work effectively, with new entrants like Hyo, Clling, and OpenAI's Sora making significant strides in the market.

We recently raised $17 million from Lightseed and Nvidia, and our revenue is doubling every quarter as we build the next generation of search.

OpenAI's recent $6.6 billion funding round indicates that AI models are becoming increasingly capital intensive, with future models expected to be orders of magnitude larger than current ones.

AI-generated video and audio have been consistently funded by Y Combinator since 2016, and the quality of these technologies is improving.

The introduction of Llama 2 activated a wave of AI development across various sectors, enabling companies and researchers to engage with AI more effectively.