You upscale a clip, watch it full screen, and it looks great. Then you zoom to 100% to check a face, and it looks worse than the original. Waxy skin, painted edges, patterns where there was noise.

This is not a bug in the tool. It is the difference between resolution and detail, and it is measurable.

Resolution is not detail

A neural upscaler does not invent detail. It predicts, for every new pixel, the most probable value given its neighborhood.

At 2x, one pixel becomes four. Three are predicted. The ratio of measured to predicted information is still workable.

At 4x, one pixel becomes sixteen. Fifteen are predicted. Beyond that point the model stops looking at your image and starts reproducing patterns it memorized during training.

That is where the artifacts come from: uniform texture where there was grain, glassy skin, edges that look drawn rather than photographed.

before and after comparison of a portrait upscaled to 4K - Natural HD mode keeps skin texture
Preset NATURAL HD. Left is the blurry original, right is the result. Skin texture is preserved.

Why full screen lies to you

Your eye cannot separate resolution from detail. A larger image always reads as sharper at a glance, because there is simply more signal reaching the retina.

At 100%, that advantage disappears. You see one screen pixel for one image pixel, and the prediction errors become visible.

So the only valid test is: same crop, same zoom, side by side, at 100%. Anything else is measuring file size, not quality.

The four-step order that actually works

Most people get the order wrong, then blame the model.

Step 1: evaluate the damage. Open at 100%. If the noise is compression (8x8 blocks), you need cleanup first. If it is film grain, aggressive cleanup will erase the original texture along with it.

Step 2: denoise, gently. Always less than feels necessary. Whatever noise remains is part of the document; the noise the algorithm adds is not.

before and after removing compression noise from a social media video - Social Denoise mode
Preset SOCIAL DENOISE. Left is the re-downloaded clip, right after cleanup. Block artifacts are gone.

Step 3: upscale 2x, not 4x. From 720p to 1440p. Going to 4K from 720p means you are asking the model to invent 93% of the pixels.

Step 4: sharpen last, and lightly. Sharpening before upscaling means the algorithm amplifies the halos you just created.

before and after on a building facade - Clear Reality mode removes compression noise, keeps detail
Preset CLEAR REALITY. Structural detail survives; only compression noise is removed.

Three things to look at when judging

Not the face alone. Check these:

  • Hair and eyelashes — must separate into strands, not merge into a mass
  • Fabric weave — the pattern must keep its periodicity
  • Object edges against background — clean, no grey halo

If hair looks right but fabric turned into a flat patch, the mode was wrong, not the strength.

before and after on fabric and wood texture - True Texture mode keeps the pattern regular
Preset TRUE TEXTURE. The repetition should stay regular, with no stretched or compressed areas.

Where to test it

I use UpgoLAB in the browser. Modes are separated by material type, and the denoise, upscale, sharpen order is already wired internally — which is the exact part most people get wrong by hand. The mode list shows which one is meant for which material. Free to try, no install.

What does not work

Being straight about the limits:

  • Information that was never captured does not come back. Fully blurred text can become legible, but not correct.
  • Motion blur is not recoverable. Direction information is gone at capture time.
  • Heavily compressed material has a low ceiling. High-frequency detail was already discarded.

These tools extract the maximum from what exists. They do not create new material.

Summary

  • 100% zoom is the only honest test
  • Denoise first, upscale second, sharpen last
  • Stay at 2x
  • Pick the mode by material, not by habit

The order of steps changes the result more than any single setting.