Moescape Diffusion is an advanced tag-based AI image generation model specialized in creating high-quality anime-style artwork. Moescape Diffusion 1.0, its latest flagship version, is built on the robust SDXL architecture with a massive dataset of 8M diverse anime-style images and trained for approximately 2,500 GPU hours with an extraordinary exposure to over 80M training images, this exclusive model implements well-known tag ordering systems to deliver exceptional quality and control in anime art generation.
The model offers two specialized variants:
- Full Version: designed for power users requiring granular control and flexible outputs
- Aesthetic(sft) Version: streamlined for end users seeking quick, beautiful generations
Description
Moescape Diffusion 1.0 Aesthetic is a stylized finetuned version of Moescape Diffusion 1.0, optimized for quick, beautiful generations. This version simplifies the creation process while maintaining high aesthetic quality. Perfect for end users seeking rapid, visually appealing results without complex prompt engineering. While it may have some limitations in anatomy accuracy and tendency toward synthetic styles, it excels in producing consistently attractive outputs.
Usage Guide:
- Use tags instead of natural language
- Follow this structured prompt template for optimal results: `1girl/1boy, character name, series source, artist reference, followed by additional tags in any order`
- Special Tags:
1. Quality tags: masterpiece, best quality, low quality, worst quality
2. Score tags: high score, great score, good score, average score, bad score, low score
3. Temporal tags: year 2005, year 2015, year 2020, ... year n
4. Rating tags: safe, sensitive, nsfw, explicit
Recommended settings:
- For higher quality outputs, append these tags at the start or at the end of your prompt: `masterpiece, high score, great score, absurdres`. Note: Unlike other models, score tags have more impact than quality tags in steering output quality
- Use the recommended negative prompt
- Sampler steps: 20-28 (25 recommended)
- CFG scale: 4-7 (5 recommended)
- Preferred sampler: Euler Ancestral (Euler A)
- Supported Resolution : 1024 x 1024, 1152 x 896, 896 x 1152, 1216 x 832, 832 x 1216, 1344 x 768, 768 x 1344, 1536 x 640, 640 x 1536
- Suggested style tags: faux traditional media, watercolor, oil painting, impressionism, impasto, abstract, chiaroscuro, fine art parody, art nouveau, etc
- Example Prompt Structure: 1girl, firefly \(honkai: star rail\), honkai \(series\), honkai: star rail, casual, solo, looking at viewer, outdoors, smile, reaching towards viewer, night, masterpiece, high score, great score, absurdres