Back to blog

Increasing resolution without losing quality: what is possible

How many megapixels you need for each print size, how far a 2x or 4x AI upscale really gets you, and when upscaling is not the fix at all.

Aug 31, 2026Mejorar TeamMejorar Team

"Increase the resolution without losing quality" sounds like a contradiction, and for years it was: enlarging an image only made its pixels bigger. Today's methods reconstruct plausible detail instead of stretching it, and the jump is real — but it has limits worth knowing before you send anything to print.

How much resolution you actually need

The working rule for print is 300 dpi (dots per inch). That is where the numbers come from:

| Print size | Pixels needed | Megapixels | | ------------------------------- | ------------- | ---------- | | 4 × 6 in (10 × 15 cm) | 1200 × 1800 | 2.2 MP | | 8 × 12 in (20 × 30 cm) | 2400 × 3600 | 8.6 MP | | 12 × 18 in (30 × 45 cm) | 3500 × 5300 | 18 MP | | Poster, 20 × 28 in (50 × 70 cm) | 5900 × 8300 | 49 MP |

With one important caveat: the bigger the print, the further away it is viewed. A poster seen from two meters tolerates 150 dpi without anyone noticing. The 300 dpi figure is for what you hold in your hand.

For screens the bar is far lower: a photo 2000 px wide looks perfect on almost any monitor and on social media.

What a 2× or 4× upscale really does

An upscaling model does not "find" hidden detail: it predicts what should be there, based on millions of real images. On natural textures — skin, fabric, leaves, brick — it is convincing. On detail that carries meaning — letters, numbers, specific features — it can invent.

What to expect:

  • is the safe jump. It almost always improves things without odd artifacts.
  • works well if the original is already decent. On a poor original, it multiplies the flaws too.
  • Upscaling twice in a row (4× and then another 4×) does not give you 16×: it gives an increasingly "painted" result, because the second pass works on the first one's inventions.

The most common mistake: upscaling when the problem is something else

A lot of people upscale a photo that is really blurry or noisy. Upscaling then just produces a larger version of the same problem.

The right order when there are several problems:

  1. Remove the noise, if the photo is dark or grainy.
  2. Remove the blur or improve sharpness.
  3. Increase the resolution last, on a clean base.

Upscaling is always the last step, never the first.

Before you print

  • Check the result at 100%, not as a thumbnail.
  • Look closely at text and faces: those are the areas where an invented reconstruction shows first.
  • Save the upscaled file separately. Do not overwrite the original: it is the only copy that still holds the real information.