Proximasanfinetuning
Models by this creator
🤷
fantassified_icons_v2
51
The fantassified_icons_v2 model is a new and improved version of the previous fantassified_icons model, created by the maintainer proximasanfinetuning. This model generates icons inspired by fantasy games, with mostly plain backgrounds. It was trained on a dataset consisting mostly of the old version's dataset, but the maintainer has learned some new things since the dreambooth days. The model is comparable to similar icon generation models like the kawaiinimal-icons model, which generates cute animal-themed icons, and the IconsMI-AppIconsModelforSD model, which is aimed at generating high-quality app icons. Model inputs and outputs Inputs Text prompts that describe the desired fantasy-themed icon, such as "a lemon themed high quality hamburger" Outputs Realistic, high-quality images of fantasy-themed icons matching the provided prompt The model can generate multiple images per prompt (e.g. 6 images) Capabilities The fantassified_icons_v2 model is able to generate a wide variety of fantasy-themed icons, from potions and magical items to creatures and fantastical landscapes. The examples provided show a good range of what the model can produce, including animated icons, simple icons with plain backgrounds, and more detailed icons. What can I use it for? This model could be useful for game developers, app designers, or anyone looking to create fantasy-themed icons or illustrations. The maintainer notes that it may not work as well for generating images of people or faces, as those were not a focus during training, but it should work well for items, creatures, and other fantasy elements. Things to try One interesting thing to try with this model is using it to generate icons for a fantasy-themed app or game. The simple backgrounds and focus on items and creatures could work well for mobile app icons, in-game UI elements, or other graphical assets. You could also experiment with different prompts and prompt engineering techniques to see what kinds of fantastical icons the model can produce.
Updated 8/29/2024
📉
luna-diffusion
45
luna-diffusion is a fine-tuned version of Stable Diffusion 1.5 created by proximacentaurib. It was trained on a few hundred mostly hand-captioned high-resolution images to produce an ethereal, painterly aesthetic. Similar models include Dreamlike Diffusion 1.0, which is also a fine-tuned version of Stable Diffusion, and Hitokomoru Diffusion, which has been fine-tuned on Japanese artwork. Model inputs and outputs luna-diffusion is a text-to-image generation model that takes a text prompt as input and produces an image as output. The model was fine-tuned on high-resolution images, so it works best at 768x768, 512x768, or 768x512 pixel resolutions. The model also supports adding "painting" to the prompt to increase the painterly effect, and "illustration" to get more vector art-style images. Inputs Text prompt**: A natural language description of the desired image, such as "painting of a beautiful woman with red hair, 8k, high quality" Outputs Image**: A generated image matching the provided text prompt, saved as a JPEG or PNG file Capabilities luna-diffusion can generate high-quality, painterly-style images based on text prompts. The model produces ethereal, soft-focus images with a focus on detailed scenes and figures. It works particularly well for prompts involving people, nature, and fantasy elements. What can I use it for? luna-diffusion is well-suited for applications in art, design, and creative expression. You could use it to generate concept art, illustrations, or other visual assets for things like games, books, marketing materials, and more. The model's unique aesthetic could also make it useful for mood boards, visual inspiration, or other creative projects. Things to try To get the best results from luna-diffusion, try experimenting with different aspect ratios and resolutions. The model was trained on 768x768 images, so that size or similar ratios like 512x768 or 768x512 tend to work well. You can also play with the "painting" and "illustration" keywords in your prompts to adjust the style. Additionally, the DPM++ 2M sampler often produces crisp, clear results, while the Euler_a sampler gives a softer look.
Updated 9/6/2024