Anmokomergetest1

🚀 Anmokomergetest1

Anmokomergetest1 is a merged Stable Diffusion model that specializes in anime. It can generate vivid, detailed, and high - quality anime pictures across a wide range even with simple descriptions.

It is an authorized copy of @路费不够's private model from the Chinese AI community NovelAI 中文频道. This model is merged from three models: Anything - V3.0(Linaqruf), momoko - e(anonymous), and mergetest1 (another private model from the community).

It performs better at 1k original resolution. With Hires.fix, it can produce larger pictures with more logical structures.

The model is available in two versions: an uncompressed checkpoint and a compressed safetensors. The CKPT needs to work with an individual VAE; otherwise, the output won't be vivid. It is better for merging. The safetensors has a baked - in VAE and is better for generating pictures.

⚠️ Important Note

No commercial usage!

✨ Features

• Specializes in generating high - quality anime pictures.
• Can generate vivid and detailed images with simple descriptions.
• Available in two versions for different use - cases (merging and image generation).
• Performs well at 1k original resolution and can produce larger pictures with Hires.fix.

📚 Documentation

Preview

• Hatsune Miku, upper body
  
• Kagamine Rin, upper body
  
• Hakurei Reimu, upper body
  
• Tohsaka Rin, upper body
  
• Misaka Mikoto, upper body
  
• Yuuki Asuna, upper body
  

Usage

Use it as a normal Stable Diffusion model package v1.x, and no external yaml config is needed.

💡 Usage Tip

Recommended settings: Steps: 12 - 28, Sampler: DPM++ SDE Karras, CFG scale: 5 - 11, Resolution: 1024x1024

Tags

For both positive and negative tags, less is more.

💡 Usage Tip

Recommended Positives for Quality Control: masterpiece, best quality. You can add other prompts as you like.

Recommended Negatives: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, artists name

Bias

• Low quality at too low or too high resolutions!!!
• At low resolutions, the eyes may look bad, and malformed limbs often appear.
• The training set contains copyright images.

Formula

• Round 1: Anything - V3.0(Linaqruf) + momoko - e(anonymous), sum rate 0.25
• Round 2: () + mergetest1 (Private, from the Chinese AI community NovelAI 中文频道), sum rate 0.5

We get Anmokomergetest1.ckpt, which is better for further merges. After baking in the vae - ft - mse - 840000 - ema - pruned(StabilityAI) VAE, pruning ema, and compressing to FP16, we get Anmokomergetest1.safetensors, which is better for generating images.

📄 License

This project is licensed under the CC - BY - NC - 4.0 license.