Why 4× Is the Most Requested Upscale
Of all the enlargement factors, 4× is the one people actually need. It's the jump that turns a web-sized image into something usable in the physical world: a 500px thumbnail becomes 2000px, comfortably enough to print at postcard size at full 300 DPI quality. It's also the standard output of most AI super-resolution models, because it's the point where reconstruction still looks believable.
What 4× Actually Means
| Original | After 4× | Total pixels |
|---|---|---|
| 300 × 300 | 1200 × 1200 | 16× more |
| 500 × 500 | 2000 × 2000 | 16× more |
| 800 × 600 | 3200 × 2400 | 16× more |
| 1024 × 768 | 4096 × 3072 | 16× more |
Note that both dimensions multiply by four, so the area — and roughly the file size — grows sixteenfold. A 200 KB source can easily become a 2–3 MB output. That's expected; compress the result afterwards if you need it smaller.
Fast vs AI at 4×
At 2× the difference between engines is modest. At 4× it becomes obvious: classical scaling has to invent fifteen out of every sixteen pixels, so edges go soft no matter how much you sharpen. AI mode runs the ESRGAN model twice (2× then 2× again), reconstructing plausible detail at each stage — the result holds together far better on faces, text and fine texture.
The trade-off is time and size limits. AI mode processes on your own device, so a 1000×1000 input at 4× can take a minute or more. For large source images or big batches, Fast mode with Medium sharpening remains a genuinely good result.
Getting the Best 4× Result
- Start sharp. A small, crisp image upscales beautifully. A blurry one just becomes a bigger blurry one.
- Avoid compressed sources. JPEG artifacts get magnified sixteenfold along with everything else.
- Two 2× passes beat one 4× jump in Fast mode — which is exactly what this tool does automatically.
- Compress last. Upscale first, then run the output through the JPG compressor if file size matters.
Need a different factor? The main image upscaler offers 2× through 8×, and enlarge image covers custom sizing.