Magic stable diffusion is a fork from Stable Diffusion, please, read the next website to know more about the original software(README LITTLE APPRENTICE).

Magic Wonder Difussion pursues the next objectives:

• Allow the possibility to generate big images in computers with less VRAM.
• Become the AI image generation in something more accessible to all the people.

All the parameters from stable diffusion are compatible, Magic Wonder Difussion has the next own parameters:

• --resize(Default 2): The upscale factor for the resized and improved version
• --nsfw_protection(default 0): Non safe for work flag
• --watermark_protection(default 0): Watermark flag

NOTE: The output in Magic Wonder Difussion is in three folders; original, resized, and improved

The next roadmap is one estimation, some new features can be added in a different order:

• Modify img2img to allow half precision => February-March
• Add one video2img feature. => February-March-May
• Move resize code to the main aplication, magic-wonder-diffusion => May
• Prepare software to be one isolate module => May
• Add one dictionary to control prompts => Unknow
• Create one original model for Magic Wonder Diffusion => Unknow
• ;)

--prompt "An oil picture about one vintage astronaut in mars" --plms --n_iter 10 --n_samples 1 --ckpt YOUR_DRIVE:/YOUR_ROUTE/model.ckpt --ddim_steps 200 --H 576 --W 384 --resize_factor 2 --nsfw_protection 1

The next examples are related to the original, resized, and improved version

!