Luma AI

Develops Dream Machine and Ray3 model for text-to-video and image-to-video generation with HDR pipeline. Launched June 2024; Ray3.14 update adds native 1080p, 4x faster performance, and 3x lower cost.

Ray3 introduces first-to-market HDR video generation and reasoning-driven approach with character reference consistency.

lumalabs.ai Foundation Models - Video

United Statesprivate

At a glance

Total raised

$1.1B

last round series c

Headcount

101-200

approximate · public range

Headquarters

Palo Alto

United States

Founded

2021

5 years active

Status

Private

Foundation Models - Video

Latest valuation

$4B

Series C · 2025

For subscribers

Subscribe to see the full profile

Including Funding rounds, Bull / Bear thesis, Stock + earnings, Roster changes, Patents, News, and Open roles.

Founding

50% off

/month

94 of 100 spots left

Full

$10

/month

Available once all 100 Founding Member spots are claimed.

Get full access

Already subscribed? Sign in →

Funding history

Total raised

$1.1B

Rounds tracked

Latest stage

Series C

Latest valuation

$4B

Date	Stage	Amount	Valuation	Lead investors
Nov 19, 2025	Series C	$900M	$4B	HUMAIN
Dec 6, 2024	Series C	$90M	—	Amazon, AMD
Jan 9, 2024	Series B	$43M	$250M	Andreessen Horowitz
Mar 20, 2023	Series A	$20M	$100M	Amplify Partners
Oct 30, 2021	Seed	$4.3M	—	Matrix Partners

Founder backgrounds

Alberto Taiuti

Co-Founder

Cofounder and CTO with deep roots in AR/VR systems. Previously served as a Senior Autonomy Software Engineer at Skydio and worked at Apple as an AR/VR Engineer. Also worked at Codeplay Software and published Slide AR. Attended Abertay University. Co-founded Luma AI in 2021 alongside Amit Jain, both former Apple employees where both had worked on computer vision and AR technologies. His background in real-time graphics, autonomous systems, and spatial computing directly informed the company's pivot from 3D capture to video generation—extending visual fidelity expertise into generative modeling.

Alex Yu

Co-Founder and CTO

Graduated from UC Berkeley where he conducted early research on NeRF neural rendering under Professor Angjoo Kanazawa, with internships at Google and Adobe. After graduating in 2021, he turned down PhD offers from Stanford and MIT to start Luma AI. Authored pioneering papers including pixelNeRF (CVPR 2021) and PlenOctrees for Real-time Rendering of Neural Radiance Fields (ICCV 2021), which achieved real-time 3D neural rendering at scale. Founded Luma with CEO Amit Jain after recognizing that NeRF technology solved long-standing 3D reconstruction challenges that Jain had encountered during Apple Vision Pro development. His leadership philosophy emphasizes product quality as the primary moat for consumer software, with equal focus on data and model excellence in AI.

Amit Jain

Co-Founder and CEO

Holds a BS in Mathematics and Computer Science from Missouri Valley College. Worked as a Software Engineer at Circle Medical (2015–2017) and briefly as an iOS Engineer at Product Hunt (2015), followed by roles at Apple as a Systems and Machine Learning Engineer, leading development of the Passthrough feature for the Apple Vision Pro and integrating the first LiDAR sensors into the iPhone. Founded Luma in 2022. Scaled the company from a 3D-capture pioneer to a leader in generative video, raising a $900M Series C. Believes AI can help the film industry take risks again, positioning Luma as a bridge between technical innovation and creative application.

AI bull / bear

Bull case

Luma AI's Dream Machine and Ray3 model demonstrate clear technical leadership in the competitive video generation space. The Ray3.14 update's 4x performance improvement and 3x cost reduction address critical barriers to adoption—speed and economics—that hobbled earlier text-to-video tools. Native 1080p output and first-to-market HDR pipeline position Luma to capture premium enterprise use cases in advertising, entertainment, and content creation where quality justifies pricing power. The reasoning-driven approach with character reference consistency solves a persistent pain point: maintaining visual coherence across generated scenes. This capability is essential for narrative applications and signals differentiated IP beyond simply scaling transformers. Launching in June 2024 and iterating rapidly to Ray3.14 shows execution velocity in a market where model obsolescence is measured in quarters. If Luma can maintain this technical edge and translate it into sticky enterprise contracts before OpenAI's Sora or Runway scale further, it carves out defensible margin in the generative video stack.

Bear case

Luma AI operates in a brutally competitive and capital-intensive foundation model race with undisclosed funding—a red flag given rivals like Runway (Series C, reported $141M total) and well-capitalized incumbents (OpenAI, Google, Meta) pouring billions into video generation. Without transparent backing, Luma risks being unable to sustain the compute and talent arms race required to stay relevant as frontier labs commoditize video capabilities. The model's June 2024 launch is recent, and despite technical wins like HDR and character consistency, these features may not constitute a durable moat. OpenAI's Sora and Google's Veo demonstrate that video quality gaps close rapidly. Cost and speed improvements are easily replicable by competitors with deeper pockets. Luma has no disclosed enterprise traction, revenue model, or user numbers, making it unclear whether Dream Machine can monetize before larger platforms bundle similar capabilities for free. The lack of funding visibility suggests either bootstrapping constraints or difficulty convincing investors of defensibility—both problematic in this subsector.

Generated by claude-sonnet-4-5-20250929 · May 1, 2026Refreshes quarterly · see disclosures

In the news

Comfy Org (official)5/5/2026
Luma Uni-1 is now available via Partner Nodes
Luma Uni-1 image generation model now available in ComfyUI via Partner Nodes, featuring autoregressive reasoning-first architecture with photorealism and advanced text rendering.

Watch

Leadership

Amit Jain

Co-Founder and CEO

Alex Yu

Co-Founder and CTO

Alberto Taiuti

Co-Founder

—

Caroline Ingeborn

Chief Operating Officer

in 𝕏

Products

Dream Machine
Dream Machine is Luma AI's flagship platform for cinematic video generation. It enables creatives to generate professional-grade videos and images from text, image, or video prompts. Powered by the Ray3 video model, Dream Machine supports advanced features including text-to-video, image-to-video, video modification with character references, start/end frame control, and loop creation. The platform serves over 30 million users globally and supports commercial use. It integrates with Adobe Creative Cloud and offers both web and iOS applications.
Ray3
Ray3 is Luma AI's world-first reasoning video model capable of understanding director intent and planning shots before generation. Released in September 2025, Ray3 introduces native HDR generation in 10-bit, 12-bit, and 16-bit formats, enabling studio-grade visual fidelity. Key capabilities include reasoning-driven generation, character consistency across shots, keyframe control, and realistic physics simulation. Ray3 includes Draft Mode for rapid exploration and Hi-Fi Diffusion for production-grade 4K HDR mastering. The latest version, Ray3.14 (January 2026), delivers native 1080p output at 4x faster speeds with enhanced reasoning.
Uni-1
Uni-1 is Luma AI's unified reasoning image model released in March 2026, designed for understanding, reasoning, and generation in a single architecture. It combines image understanding and generation capabilities, enabling models to evaluate and refine outputs iteratively. Uni-1 outperforms Google's Nano Banana 2 and OpenAI's GPT Image 1.5 on reasoning benchmarks while costing 10-30% less at 2K resolution. The model excels at complex compositional accuracy, multi-reference generation, and precise prompt adherence, making it ideal for production workflows requiring consistent brand quality across campaigns.
Ray3.14
Ray3.14 is Luma AI's professional-grade upgrade to Ray3, released in January 2026. It solves the quality-speed-reasoning tradeoff by delivering native 1080p output at 4x faster generation speeds with enhanced reasoning capabilities. Designed for creators and studios requiring production-ready results, Ray3.14 maintains the world's first native HDR generation with studio-grade fidelity while reducing inference costs. It integrates seamlessly with existing Ray3 workflows and Adobe Creative Cloud, supporting text-to-video, image-to-video, video-to-video, and video enhancement at production scale.

Open roles

We don't have a live feed for this company's ATS. Their careers page has every open role.

View all careers ↗