🎨 AI Image Generator Pro - NSFW Capable
Advanced Image Generation with ControlNet, LoRA & VAE Support
⚠️ Content Warning: This tool can generate NSFW content. Use responsibly and in compliance with applicable laws.
⚠️ Running on CPU - Generation will be slower
Transform sketches/images using SD1.5 with ControlNet
- Supports: lineart, lineart_anime, canny, depth, normal, openpose, softedge, segmentation, mlsd, shuffle, scribble, tile
- Best Resolution: 512x512
Model Settings
SD1.5 Base Model
ControlNet Type
Enhancement Options
LoRA Model
0.1 2
VAE Model (Optional)
Generation Parameters
10 100
1 30
0.1 2
Transform sketches/images using SDXL with ControlNet
- Supports: canny_sdxl, depth_sdxl, openpose_sdxl, lineart_sdxl (new!)
- Best Resolution: 1024x1024
- Higher quality, more VRAM required
Model Settings
SDXL Base Model
ControlNet Type
Enhancement Options
LoRA Model
0.1 2
VAE Model (Optional)
Generation Parameters
10 100
1 30
0.1 2
Specialized Anime Lineart Colorization
- Model: ShinoharaHare/Waifu-Colorize-XL
- Specialized for: Anime/manga lineart coloring
- Features: Automatic colorization with vibrant anime colors
- Best Resolution: 1024x1024
Model Settings
Using: ShinoharaHare/Waifu-Colorize-XL (Specialized Anime Colorization)
Enhancement Options
0.1 2
VAE Model (Optional)
Generation Parameters
10 50
1 20
0.1 2
Tips for Best Results:
- Use clean lineart for best results
- Higher ControlNet weight (1.0-1.5) for better line following
- Lower CFG scale (5-8) for more natural coloring
- Add color hints in prompt (e.g., "blue hair, red eyes, pink dress")
- Keep prompts simple for this specialized model
Generate images from text descriptions using SD1.5
- Best Resolution: 512x512, 512x768, 768x512
- Faster generation, lower VRAM usage
Model Configuration
SD1.5 Base Model
Enhancement Options
LoRA Model
0.1 2
VAE Model
Prompts
Generation Parameters
10 100
1 30
256 1024
256 1024
Ultra-Fast Image Generation with SDXL-Turbo
- Model: stabilityai/sdxl-turbo
- Features: 1-4 steps generation, extremely fast
- Best Resolution: 512x512 to 1024x1024
- Warning: Lower quality than full SDXL but much faster
Model Configuration
Turbo Model
Enhancement Options
LoRA Model
0.1 2
VAE Model
Prompts
Generation Parameters (Turbo needs only 1-4 steps!)
1 10
0 10
256 1024
256 1024
Turbo Model Tips:
- Use very few steps: 1-4 steps is enough!
- Low CFG Scale: 0.0-2.0 works best
- Fast but lower quality: For quick previews/testing
- Works well with: Simple prompts, concept testing
Generate images from text descriptions using SDXL
- Best Resolution: 1024x1024, 1024x1536, 1536x1024
- Higher quality, more detail, better composition
Model Configuration
SDXL Base Model
Enhancement Options
LoRA Model
0.1 2
VAE Model
Prompts
Generation Parameters
10 100
1 30
512 2048
512 2048
Model & Feature Guide
🎯 Model Comparison
SD1.5 (Stable Diffusion 1.5)
- Pros: Fast, low VRAM, many models
- Cons: 512px max, lower quality
- Best for: Quick tests, anime, lower-end hardware
SDXL (Stable Diffusion XL)
- Pros: 1024px+, high quality, better composition
- Cons: High VRAM, slower
- Best for: Final quality, professional work
SDXL-Turbo
- Pros: Extremely fast (1-4 steps!)
- Cons: Lower quality than full SDXL
- Best for: Quick previews, concept testing
Waifu-Colorize-XL
- Pros: Specialized for anime lineart coloring
- Cons: Anime-only, requires clean lineart
- Best for: Anime/manga colorization
🎨 New ControlNet for SDXL
Lineart ControlNet for SDXL
- Model: ShermanG/ControlNet-Standard-Lineart-for-SDXL
- Purpose: Convert lineart to colored images
- Best with: Clean lineart, anime/manga styles
- Used in: Waifu-Colorize-XL Tab
💎 Recommended Workflows
For Anime/Manga Artists
- Draw lineart
- Use Waifu-Colorize-XL Tab for automatic coloring
- Or use SDXL ControlNet with lineart_sdxl
For Quick Concepts
- Use SDXL-Turbo Tab for 1-4 step generation
- Refine in SDXL Text-to-Image if needed
For Professional Work
- Use SDXL Text-to-Image with Refiner
- High steps (40-50), CFG 7-9
- 1024x1024 resolution
⚡ Turbo Model Tips
SDXL-Turbo Best Practices:
- Steps: 1-4 only!
- CFG Scale: 0.0-2.0
- Prompts: Keep simple
- Resolution: 512x512 to 1024x1024
- Use case: Storyboarding, concept art, quick iterations
🌸 Waifu-Colorize-XL Tips
For Best Results:
- Clean lineart: No stray marks
- ControlNet weight: 1.0-1.5
- CFG Scale: 5-8
- Simple prompts: "vibrant colors, anime style"
- Resolution: 1024x1024
Example Workflow:
- Draw/sketch in your favorite app
- Export as clean lineart (black on white)
- Upload to Waifu-Colorize-XL Tab
- Adjust parameters as needed
- Generate and refine
🚀 Performance Optimization
Low VRAM (<8GB)
- Use SD1.5 models only
- 512x512 resolution
- Enable attention slicing
Medium VRAM (8-12GB)
- SD1.5 and SDXL (no refiner)
- 1024x1024 for SDXL
- Enable xFormers
High VRAM (12GB+)
- All models including SDXL with refiner
- Higher resolutions
- Multiple LoRAs
🔄 Memory Management
When to Unload Models:
- Switching between SD1.5 and SDXL
- Getting "out of memory" errors
- Changing ControlNet types
- After long generation sessions
Memory Saving Tips:
- Use "Unload All Models" button
- Generate in batches
- Lower resolution for testing
- Close other GPU applications