Why Resolution Matters
Resolution determines the level of detail in your image. Higher resolution = more detail, but also more generation time and cost. The key is matching resolution to your use case.
Standard Resolutions by Model
| Model | Native | Best Range |
|---|---|---|
| Flux Dev/Pro | 1024x1024 | 768-2048 per side |
| SDXL | 1024x1024 | 768-1536 per side |
| SD 1.5 | 512x512 | 384-768 per side |
Important: Going below native resolution produces blurry results. Going far above wastes time without quality gain (use upscaling instead).
Common Aspect Ratios
- 1:1 (Square) — Instagram posts, profile pictures. 1024x1024.
- 3:4 (Portrait) — Pinterest, portraits, character art. 768x1024.
- 4:3 (Landscape) — Desktop wallpaper, presentations. 1024x768.
- 16:9 (Widescreen) — YouTube thumbnails, cinematic. 1216x684.
- 9:16 (Vertical) — TikTok, Instagram Stories/Reels. 576x1024.
The Upscaling Trick
Instead of generating at 2048x2048 (slow, expensive, often worse quality), generate at 1024x1024 and then upscale 2x or 4x. This is faster, cheaper, and often produces better results because the AI works best at its native resolution.
EGAKU AI has a built-in 4x upscaler (RealESRGAN) — one click to go from 1024px to 4096px.
Common Pitfalls
- Don't use odd numbers. Stick to multiples of 64 (e.g., 768, 832, 1024, 1216).
- Don't mix up width/height. 768x1024 is portrait, 1024x768 is landscape.
- Non-square ratios need model support. Flux handles any ratio well. Some SDXL/SD1.5 models are trained on specific ratios.