How to upscale 480p to 1080p on a Mac
480p to 1080p works. It is a 2.25× jump in each dimension, about five new pixels for every real one, and modern models handle that convincingly enough that the result usually looks like it was shot on better equipment. 480p to 4K is 27 new pixels per real one, and that is where reconstruction turns into invention.
Why 1080p is the honest target
| From 640×480 | Pixels to invent, each real one | Realistic? |
|---|---|---|
| → 1280×720 | 3 | Comfortably |
| → 1920×1080 | ~5 | Yes — the sweet spot |
| → 3840×2160 | 27 | Diminishing; faces suffer |
The failure at 4K is not that it looks blurry — it is that it looks confident and wrong. Skin smooths into plastic, fabric weave becomes a repeating pattern the camera never saw, and small text becomes crisp nonsense. At 1080p the model has enough real signal to stay anchored.
There is also a practical argument. Most people watching your footage are on a laptop or a phone, where 1080p is already at or beyond the panel's ability to show a difference, and the file is a quarter the size.
Do these two things first
Denoise before upscaling. 480p sources are usually old, and old usually means noisy. The model cannot tell noise from detail, so it will reconstruct the grain into permanent texture — and once it has, no later denoise removes it, because it is no longer noise.
Deinterlace if it came off tape or a DVD. Comb teeth on moving edges get sharpened into permanent artefacts by any upscaler.
One pass, not two
Do not go 480p → 720p today and 720p → 1080p next week. The second pass treats the first pass's invented detail as real source material and builds on top of it, compounding whatever it got wrong. Pick the final resolution and get there in a single operation.
Judging the result properly
Watch it playing, at full size. Not paused, and not in a small preview pane. Frames are upscaled independently, so the model's guesses differ slightly frame to frame, and in flat areas — a wall, the sky, a plain shirt — that difference reads as a fine crawling shimmer. It is invisible in a screenshot and obvious in motion, which is why still-frame comparisons of upscalers tell you almost nothing.
If you see it, that is the tool, not your footage. It is the axis good upscalers win on and the one most reviews never test.
In Crisp
Drop the file in, pick 2× rather than the maximum, and export. It runs on your Mac — no upload, which for a 480p file of somebody's family is the sensible default anyway. The free tier watermarks the output, so you can check the result on your own footage before deciding whether the $19 is worth it.
Where the 480p came from changes what's possible
A 480p file that was shot at 480p — a MiniDV camcorder, an early digital camera — holds close to 480 lines of real detail, and upscaling has genuine information to work from. A 480p file that was downscaled from something larger, or that arrived through a messaging app, is a different case: it has been through an encoder that threw away high-frequency detail deliberately, and what is left is both smaller and damaged.
Counterintuitively the second one often improves more, because compression artefacts are highly learnable — blocking and banding have consistent shapes that a model has seen a million times — while genuinely missing fine detail has to be invented from nothing. If your source came off WhatsApp or a re-uploaded video, expect a bigger jump than the resolution alone suggests.
Frame rate is worth checking too
A lot of 480p material is 29.97fps rather than a round 30, a legacy of NTSC broadcast timing. Treat it as 30 anywhere in the chain and the audio drifts progressively out of sync — inaudible at the start and half a second out by the end of a long clip. Any decent tool preserves the original rate exactly; it is worth confirming the output matches the input rather than assuming.