The new module *ai img2img* generates new images based on a flame fractal using *AI*.
It does this by using the API of the "_AUTOMATIC1111_ *StableDiffusion WebUI* ( https://github.com/AUTOMATIC1111/stable-diffusion-webui ).
The module generates prompts based on several attributes of the current flame and feeds the *AI* both with the generated prompt and a rendered image of the fractal.
There are also some options to control the creative process of the *AI*.
Together with the random flame generators this module is a literally endless image generator.
And because the module is non-modal, you may easily try out a lot of flames without closing/re-opening windows.
For example, you can just keep it open while editing you flame and create new *AI* image at any time, taking the current state of the edited flame into account.
Complete changelog:
VERSION 0.88 (20.04.2025):
added a feedback loop in the "ai img2img"-module. It feeds back parts of an generated image into the ai to further refine the result. This takes longger time, but can help to generate really awesome results
- ai img2img: new "feedback loops" option
- ai img2img: further optimizations to the prompt generator
- made the "ai img2img" an independent module. You may change the flame or load/generater a new flame in the editor without closing/re-opening windows
- refined the prompt-generating-algorithm in the "ai img2img"-module
VERSION 0.87 (17.04.2025):
ai img2img: there are now two editable prompts: "main prompt" and "modifying prompt" for better controls of the final result
- ai img2img:: added a comprehensive manual entry
- the "ai img2img"-module now uses a separate sub-folder to store the generated images. This makes it easier to keep track of all your images and config files
- added a manual entry for the gradient library
- adjusted the brightness of several random flame-generators
VERSION 0.86 (15.04.2025):
ai img2img: automatically save the original flame-image, the flame and current "ai img2img"-configuration (as text file) after generating an image. So you can come back later and refine the ai image.
- ai img2img: new upscale option
- ai img2img: new options "automatic1111 upscale script" and "automatic1111 upscale script args" in the application settings
- adjusted the brightness of several random flame-generators
- added a manual entry for the gradient editor
VERSION 0.85 (13.04.2025):
the "ai img2img"-module now got the feature add random structures into the
generated image. It does this by generating random phrases in the prompt.
- ai img2img: added more nouns to the random phrase generator
- ai img2img: added verbs to the random phrase generator
- ai img2img: added adjectives to the random phrase generator
- ai img2img: major refactoring (separating the http-stuff from the main application)
- ai img2img: choose random random-seed at first startup
VERSION 0.84 (12.04.2025):
created new module "ai img2img" which generates new images based on a flame fractal using AI.
It does this by using the using the API of the "AUTOMATIC1111 StableDiffusion WebUI ( https://github.com/AUTOMATIC1111/stable-diffusion-webui ).
It generates prompts based on the current flame, and feeds the ai both with the prompt and a rendered image.
There are also some options to control the creative process of the ai.
- ai img2img: retrieve available models from automatic1111 and display them
- ai img2img: make the desired model changeable from swan
- ai img2img: make the generation process interruptable
- ai img2img: option to select output image size
- ai img2img: option to specify "ai crerativity" (denosiing strength)
- ai img2img: option to specify cfg scale (classifier-free guidance scale)
- ai img2img: refined the promt generator (takes several attributes of the flame into the account)
- ai img2img: refined the negative prompt
- added a new setting "automatic1111 api" to the application settings to be able to change host and/or port your needs