Image to image (img2img)

Image to image, often shortened to img2img, is a generation mode that takes an existing picture as a starting point instead of generating purely from a text prompt. The model keeps the overall structure and composition of the source image while applying the changes described in the prompt, such as a new style, lighting or subject detail.

A strength setting controls how closely the result follows the original: low values make small tweaks, high values let the model reinterpret the image more freely.

In tendre.AI, img2img lets you turn a sketch, photo or earlier generation into a new variation, all processed locally on your GPU.