Luma AI Releases Ray2 Flash AI Video Model With Faster Generation Time

Luma AI, a Palo Alto-based artificial intelligence (AI) firm, released a new video generation AI model last week. Dubbed Ray2 Flash, the model is a distilled version of the recently introduced Ray 2 frontier video generation model. The company stated that the Flash model is both faster and cheaper than the large language model (LLM), while retaining the realistic video generation capabilities. The Ray2 Flash video generation model is multimodal and supports both text and images as input. It can be accessed via the Dream Machine platform that was introduced last year.

Luma AI Introduces Ray2 Flash AI Model

In a post on X (formerly known as Twitter), the official handle of Luma AI announced the launch of the new video model. The Ray2 Flash AI model is currently available within Dream Machine, and the application programming interface is expected to arrive soon.

The post claimed that the distilled video generation model is three times faster and three times cheaper than the frontier Ray2. It is also available without rate limits, meaning users can generate more videos with the same number of credits. Ray2 Flash is only available to the paid subscribers of Dream Machine.

Notably, users on the free tier get 400 one-time credits. After exhausting these, users will have to purchase a subscription. The membership to the platform starts at $6.99 (roughly Rs. 609) per month, which offers 3,200 monthly credits.

The Ray2 Flash model is based on the frontier Ray2 model, which was released in January. At the time of launch, the AI firm said that the large-scale video model can generate realistic videos with natural and coherent motion. The model also has a high prompt adherence and can stay close to the user input.

Last year, the company introduced the Dream Machine platform. The platform offers both image and video generation capabilities and users can create outputs in a wide range of styles including cinematic, animation, realistic, and more. Gadgets 360 tried the platform and found that the prompt adherence of the platform is sub-par. It struggles with multiple characters or when the prompt is too complex.