Luma opens Uni-1.1 image model API
Luma has opened a REST API for Uni-1.1, its single-model approach to image generation and editing, so developers can now build on the model the company launched a month ago.
Luma announced API access to Uni-1.1, a REST interface to its "Unified Intelligence" model for image generation and natural-language editing.
Uni-1, which launched last month, is a decoder-only autoregressive transformer. Text and image tokens share a single sequence, as both input and output, in the same forward pass. Most systems instead pair a separate language model and image model at inference time. Luma says its design gives tighter adherence to briefs with several constraints and better grounding in reference images. It also says Uni-1 was trained with input from cinematographers, VFX artists and other artists.
The API has two main endpoints. Generate Image covers text-to-image and reference-guided generation, with up to nine reference inputs per request. Modify Image handles edits described in plain language, such as changing backgrounds or lighting. Python and JavaScript/TypeScript SDKs are available, and output comes as PNG or JPEG in standard aspect ratios. Luma puts generation time at about 31 seconds per image and claims "less than half the price and latency of comparable models." It did not publish per-image pricing or name the models it compared against.
The model comes in Uni-1.1 and Uni-1.1 Max versions. Access is sold in two tiers: Build, with usage-based billing, and Scale, with higher rate limits and dedicated support.
Luma says it ranks among the top three labs in the Image Arena for text-to-image and image editing. It also claims first place on Human Preference Elo and the lead on RISEBench reasoning. Nobody has independently checked these figures. The company lists Envato, Comfy, Runware, Flora, Krea, Magnific, Fal and LovArt as having production commitments.