Stable Diffusion & ComfyUI Upscale Workflow: Best Settings
Complete guide to upscaling Stable Diffusion outputs. Compare built-in img2img, ControlNet tile, and external AI upscalers. Includes ComfyUI workflow nodes.
The Problem: SD Output Resolution
Stable Diffusion XL generates at 1024x1024. That's fine for social sharing, but too small for:
Method 1: Built-in img2img Upscale
The simplest approach — use SD's own img2img at higher resolution.
Pros
Cons
Best Settings
Method 2: ControlNet Tile Upscale
More sophisticated — uses ControlNet's tile model to maintain consistency.
Pros
Cons
Method 3: External AI Upscaler (Recommended)
Use a dedicated upscaling model that's trained specifically for resolution enhancement.
Pros
Cons
How to Use
ComfyUI Workflow
For ComfyUI users, the optimal workflow:
This hybrid approach gives you the best of both worlds: clean upscaling plus optional AI detail enhancement.
Quality Comparison
| Method | Speed | Quality | Style Preservation | Max Scale |
|---|---|---|---|---|
| img2img | Slow | Good | Medium | 2x |
| ControlNet Tile | Very Slow | Very Good | Good | 4x |
| External AI Upscaler | Fast | Excellent | Excellent | 8x |
| Hybrid (External + img2img) | Medium | Best | Very Good | 8x |
Our Recommendation
For most SD/ComfyUI users: generate at max native resolution, then use ImageUpscaler for clean 4x upscaling. If you want to add extra painterly detail, do a light img2img pass (denoise 0.1-0.2) on the upscaled result.
Related: Upscale Stable Diffusion | Upscale Midjourney | AI Art to 4K
Ready to Try?
Upload your image and see the AI difference in seconds. Free, no signup required.
Try ImageUpscaler Free