Candy Forest_SDXL_Anime Illustration style

Model Introduction
This model is trained on a dataset including direct outputs of Candy Forest and some Niji illustrations from previous LoRA training, specializing in generating various anime-style illustrations.

✅ v1.1 Fine-tuned the original training dataset at a learning rate of 1e-5, achieving better limb expression and increased background complexity.

SD1.5 Candy Forest Portal: Click me 👈

Parameter Recommendations

• CLIP Skip: 2
• VAE: Use the built-in VAE (auto-matching)
• Sampler: Euler a
• Sampling Steps: 25–35
  
• Prompt Tips
  Since Animagine implements aesthetic grading, add masterpiece, best quality, very aesthetic, absurdres to positive prompts. Add nsfw, photo, deformed, black and white, realism, disfigured, low contrast, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, duplicates, tight to negative prompts.
  
• Resolution Suggestions
  
• Direct generation: 1024×1024 or 1024×1536 (add solo for single-character scenes)
• High-definition restoration: Scale 1.5–2× based on VRAM (upscaling algorithms: R-ESRGAN 4x+ Anime6B or 4x-UltraSharp)