AI agents for creative work that generate, understand, and operate in the physical world
Generative video models for animation
Stage
Established
Not given
Founded
2021
Not given
Based in
US
US
Pricing model
Freemium
Not given
Pricing
Individual plans: Plus $30/month ($300/year) with 10,000 credits, Pro $90/month ($900/year) with 40,000 credits, Ultra $300/month ($3,000/year) with 150,000 credits. Business plans available with team management, SSO, and analytics. Video generation costs range from 20 credits/5 sec at 360p to 1200
Not given
Key features
Multimodal creative workspace for image, video, audio, and text
Ray3.2 video generation model with direct frame control and cinematic quality
Uni-1 brand intelligence model for consistent visual generation
Brainstorm Mode for conceptual planning without media generation
Create Mode for generating images, video, audio, and text
Team collaboration with shared credits and project organization
Luma Skills for building and reusing creative workflows
Export capabilities including EXR files for professional workflows
Video localization with natural voiceovers and synced visuals
Automatic model selection by Luma Agent or manual model choice
Pyramid-JiT (P-JiT) architecture with 32×32 token reduction
4× more aggressive token reduction than Linum v2
Converges in 11× fewer samples than Linum v2
2B-parameter model with frozen Qwen text encoder
Pixel-space diffusion with multi-resolution readouts
PixelREPA alignment with DINOv3-L
Curriculum learning over noise schedule
What makes it different
End-to-end creative workflows with shared context across all formats and team members
Frontier research models built in-house and shipped directly to creative professionals
Brand intelligence at the model level that maintains visual consistency
Parallel collaboration allowing teams and agents to work simultaneously
Physical AI and robotics research through Open Physical AI Lab
Low-cost training: full 2B model trained for just $21K in compute
Controls baked into model training rather than post-hoc fine-tuning
Pixel-space compression without separate VAE bottleneck
Apache 2.0 open release with full model weights and code
Luma
Linum