The latest models for generating images with artificial intelligence have very high hardware requirements, but in addition to improving their accuracy, they also introduced the concept of “rapid generation”, which lets us create images in a matter of milliseconds. The BlinkShot system presents itself as a real-time generator so fast that it can respond to every tweak the user makes to the prompt, and use previous images as a reference to keep refining its content.
The “classic ritual” for generating images locally with AI depends on four elements: we need a model, a compatible interface to access its functions, a graphics card with enough VRAM (read: “all”), and time. Unfortunately, the requirements of recent models took me out of the game completely, and my only option is to look for an online solution. But beyond platforms with tokens, hard limits, and monthly subscriptions, projects like BlinkShot are shaking up the web with some very interesting tricks. In this case, we’re talking about dynamic and nearly instant generation.
BlinkShot: Ultra-fast image generation as the prompt changes
BlinkShot is a creation of Together.AI, combining the Turbo mode of its endpoint with the virtues of the FLUX.1 Schnell model. Thanks to this, BlinkShot produces images in a few milliseconds, repeating the process for each change to the prompt. This is a very important feature, because it breaks with the traditional dynamic: instead of creating a giant prompt and “rolling the dice,” the user can add and remove details without difficulty, jump to different sections, and get an instant image that reflects each change.
The only advanced parameter in BlinkShot is the “Consistency Mode,” an equivalent to Img2Img that takes previous images as reference to generate a new frame. This definitely reminds us of the FastFlux.ai project, which was retired after generating over 500 million images. Now, I imagine that BlinkShot will not stay online forever, and its daily limit is about 50–60 images, but it works much better than expected, and in essence, it is what we want to see in terms of image generation: high speed, and immediate response to prompts.
Official site: Click here
