On this page
Luma AI makes software that turns words and pictures into video. Its Dream Machine tool went viral in the summer of 2024. Its Ray models are now built into products used by film studios, ad agencies and brand teams. The company describes itself as a lean team in Palo Alto. It is one of several AI video companies that have grown up within a few miles of University Avenue.
This profile covers what Luma makes, how it started, how it was funded and how it relates to Palo Alto. Luma does not publish a street address, so there is nothing to visit. If you are interested in working there, its careers page is the place to start.
What Luma makes
Luma's products sit on top of models it trains itself. The company says this is the point. In its own words, speed like its tools offer does not come from a wrapper around someone else's model.
- Luma Agents. The main product in 2026. Agents research, generate and refine work across video, image, audio and text for creative teams. Wikipedia notes that from September 2026 new users sign up through Luma Agents.
- Ray. The video model family. Ray3 arrived in 2025 and was billed as a video reasoning model with 16-bit HDR output. Ray3.2, released with an API in June 2026, focuses on control over each frame and continuity across cuts.
- Uni-1. A model the company calls brand intelligence. It carries style and character references into each asset. An API for Uni-1.1 launched in May 2026.
- Photon. A family of image models aimed at high quality output at low cost.
- API access. Developers can call Luma's models directly and pay by usage.
Luma also hosts some models from other labs inside its tools. In October 2026 it added Kling 4.0, for example. Its website mixes product news with a stream of how-to articles for marketers and filmmakers.
Founding and early work
Luma was founded in 2021. Amit Jain is co-founder and chief executive. Alex Yu is the other co-founder named in company profiles. Coverage of the company's 2025 funding noted that its team includes inventors of two well known techniques. NeRF turns ordinary photos into 3D scenes. DDIM speeds up image generation.
That background shaped Luma's first products. The company began with 3D capture, letting people turn phone video into 3D scenes. It then released Genie, which generated 3D objects from text prompts. Those tools found an audience among designers and game makers, but the jump to video made Luma widely known.
Dream Machine
Dream Machine launched to the public on June 12, 2024. Users typed a prompt or uploaded a still image and got a short video clip back. Early clips were five seconds long. According to the company, the tool reached one million users in four days. Social media filled with animated versions of famous paintings and memes.
The launch also drew criticism. Reviewers compared it with OpenAI's Sora and the Chinese model Kling. Several critics pointed out that Luma had not said what data it trained on, a question that hangs over the whole AI video field.
Funding and the HUMAIN deal
Luma's biggest round came in November 2025. At the U.S. and Saudi investment forum in Washington, it announced a $900 million Series C led by HUMAIN, an AI company owned by Saudi Arabia's Public Investment Fund. AMD Ventures took part. Existing investors Andreessen Horowitz, Amplify Partners and Matrix Partners joined too. CNBC reported that the round valued Luma at more than $4 billion.
The deal came with a compute partnership. HUMAIN is building a 2 gigawatt AI supercluster in Saudi Arabia called Project Halo. Luma agreed to be a customer. Luma said the money would fund research on training models with very large amounts of video and other data. It also said it planned world models for robotics, gaming and education.
Our page on Andreessen Horowitz covers one of its long-time backers. For the wider funding landscape, see AI investors on the Peninsula.
Luma timeline
| Date | Event |
|---|---|
| 2021 | Founded; early focus on 3D capture. |
| June 12, 2024 | Dream Machine opens to the public. |
| 2025 | Ray3 released; Adobe adds it to Firefly in September. |
| November 2025 | $900 million Series C led by HUMAIN. |
| December 2025 | London office announced. |
| April 2026 | Innovative Dreams, a production services company, launched with Wonder Project. |
| May and June 2026 | Uni-1.1 API, Ray3.2 and Luma Skills released. |
Luma and Palo Alto
Luma's own careers text says the team is in Palo Alto. Wikipedia and some older articles still describe the company as based in San Francisco, which is common for young AI firms that move as they grow. We follow the company's own wording.
The company is part of a cluster. Pika, another AI video company, works out of High Street downtown. Hippocratic AI, Uniphore and other AI firms have Palo Alto bases too. Our Palo Alto companies guide lists them. The draw is the same mix that has pulled startups here for decades: investors on Sand Hill Road, engineers from Stanford and the big tech campuses nearby plus a downtown full of small offices.
Stanford ties
We found no formal Stanford partnership in Luma's announcements. As with many local AI companies, the link is people. Stanford's computer science and graphics research feeds the same pool of engineers that Luma and its rivals hire from. Students interested in this field can start with the Science and Engineering Quad, where much of that work happens.
Working at Luma
Luma calls itself lean and high achieving. Its careers page lists open roles and their locations. Research, engineering and product jobs are the core, with a growing go-to-market team since the London office opened. If you are new to the area, our guide to finding a tech job in Palo Alto covers salaries, commutes and how local hiring works.
Apply only through the company's own careers page. AI startups are a frequent target of fake recruiter messages, so be wary of offers that arrive through unofficial channels.
Visiting and nearby
Luma has no public office, store or demo space. It does not publish a street address, so do not try to drop by. The best way to see its work is to try the tools online or watch the films and ads made with them.
Good to know
- No tours or walk-in visits are offered.
- Press inquiries go through the company's media page, not the office.
- Pricing and plans change often, so check lumalabs.ai before paying.
If you are in town for an interview, downtown is the easiest base. Caltrain stops at the Palo Alto station at the foot of University Avenue. Express trains reach San Francisco in about 39 minutes and San Jose Diridon in about 21. For coffee before a meeting, our Palo Alto coffee guide covers the downtown cafes. Philz on Forest Avenue and the long-running Palo Alto Creamery on Emerson Street are both a short walk from the station. For somewhere to stay, see staying downtown.
Frequently asked questions
Where is Luma AI located?
Luma AI says on its own website that it is a lean team in Palo Alto. It does not publish a street address. Some older articles describe it as San Francisco based, so check the company's careers page for the current picture. A London office was announced in December 2025.
What is Luma Dream Machine?
Dream Machine is Luma's text and image to video tool. It launched publicly on June 12, 2024 and, according to the company, reached one million users in four days. Luma's later Ray models power its video generation today.
Who founded Luma AI?
Luma AI was founded in 2021. Amit Jain is co-founder and chief executive. Alex Yu is the other co-founder named in company profiles. The team includes researchers behind NeRF and DDIM, two widely used techniques in 3D and image generation.
How much has Luma AI raised?
In November 2025 Luma announced a $900 million Series C led by HUMAIN, a Saudi AI company. AMD Ventures took part along with existing investors Andreessen Horowitz, Amplify Partners and Matrix Partners. CNBC reported a valuation of more than $4 billion.
Is Luma AI free to use?
Luma offers a free tier to try its tools and paid plans for heavier use, plus an API for developers. Plans and limits change often, so check the pricing page on lumalabs.ai before you sign up.