Sotemixv2

🚀 SoteMix V2

SoteMix V2 is a text - to - image model, trained at a high resolution, which can generate high - quality art and anime images. It offers specific settings for different resolutions to ensure optimal output.

🚀 Quick Start

SoteMix V2.1 is available here:
https://huggingface.co/Disty0/SoteMixV2.1

✨ Features

• Resolution - specific: Trained at 1024x1536, it provides better results at higher resolutions compared to 512x.
• Customizable Prompts: Allows users to use positive and negative prompts to fine - tune the generated images.
• Multiple Sampler Support: Supports different samplers like Euler a and Euler for various steps and settings.

💻 Usage Examples

Positive Prompts

(masterpiece, best quality, highres), 1girl,

Negative Prompts

(worst quality, low quality, lowres), zombie, interlocked fingers,

General Settings

• Sampler: Euler a
• Steps: 30 - 40
• Clip Skip: 1
• CFG: 7
• Base Resolution: 1024x / 768x1280 / 1280x768 / 1024x1536 / 1920x1080

Second Pass / Hires

• Sampler: Euler / Euler a
• Steps: 10 with Euler / 20 with Euler a
• Upscaler: RealESRGAN 4x+ Anime6B / ESRGAN 4x - AnimeSharp with 0.3 - 0.4 denoise strength.
• CFG: 11
• Resolution: 2x of the base resolution.

For Resolutions above 1024x1024

• Use HyperTile size 512. Smaller sizes have a very high tendency to burn the image above 2048x2048.
• Use low denoise strength and steps. Don't go above 0.3 denoise and 10 steps. Note that the model can still be chaotic at 1024x1536 and 1920x1080.

🔧 Technical Details

Training

My GPU couldn't handle full model training at these resolutions so I trained it as a 512 layer Lora with SoteMix V1 as the base. I used highres as the trigger word. Also used raifu trigger word with my OC character.

• Resolution: 1024x1536 with Bucketing
• Batch Size: 1
• Epochs: 2
• GPU: Intel ARC A770 16GB

Bucket

bucket 0: resolution (832, 1664), count: 49
bucket 1: resolution (896, 1536), count: 1
bucket 2: resolution (960, 1536), count: 12
bucket 3: resolution (960, 1600), count: 431
bucket 4: resolution (1024, 1344), count: 114
bucket 5: resolution (1024, 1408), count: 31
bucket 6: resolution (1024, 1536), count: 694
bucket 7: resolution (1152, 1344), count: 5
bucket 8: resolution (1216, 1216), count: 24
bucket 9: resolution (1344, 1024), count: 16
bucket 10: resolution (1344, 1152), count: 2
bucket 11: resolution (1536, 896), count: 55
bucket 12: resolution (1536, 1024), count: 7
bucket 13: resolution (1600, 960), count: 45
bucket 14: resolution (1664, 832), count: 21
bucket 15: resolution (2304, 576), count: 1

Merge

Merged SoteMix V1 with Lunar Radiance Light and then merged the Hires Lora I trained on top of it. Merge ratio: (0.6 SoteMix V1 + 0.4 Lunar Radiance Light) + 0.75 Hires Lora

📄 License

The model is licensed under creativeml - openrail - m.

⚠️ Important Note

Don't use this model at 512x resolutions, it will be blurry. Use SoteMix V1 for 512x instead.