# What is upscaling? How 1080p video fills a 4K screen

> What is video upscaling? Learn how 1080p video fills a 4K or 5K screen, the difference between spatial filters and AI upscaling, and what scaling can't add.

Published 2026-10-31 · Formats & codecs · https://www.hazenstudio.com/maclc/blog/what-is-upscaling/

## In short

- Video upscaling is the mathematical process of resizing a lower-resolution video frame to fit a higher-resolution physical display.
- When a 1080p video fills a 4K screen, an algorithm must calculate three new pixels for every single source pixel.
- Traditional mathematical filters interpolate between known pixels, while AI upscaling predicts likely high-frequency textures.
- Upscaling cannot recover true optical data that was never recorded or that was erased during compression.
- MacLC decodes in hardware and offers nine GPU upscaling methods in Metal—from MetalFX and FSR to neural models on 8-bit SDR video.

Video upscaling is the mathematical process of increasing the pixel dimensions of a lower-resolution video frame so that it fills a higher-resolution physical display. When you play a 1080p Full HD video full screen on a 4K monitor, an upscaling algorithm calculates three new pixels for every original pixel, expanding 2.07 million source pixels into 8.29 million display pixels.

Every non-native video you watch on a modern Mac or television undergoes upscaling. Without it, a standard 1080p video would occupy only one-quarter of a 4K screen, surrounded by wide black borders. Understanding how upscaling works, what it can and cannot achieve, and where it occurs reveals how to get the cleanest possible picture from your existing video collection.

## Why does video need to be upscaled?

Computer monitors and televisions are constructed with fixed physical grids of subpixels. A 4K Ultra HD panel always contains exactly 3840 physical columns and 2160 physical rows of pixels, totaling 8,294,400 individual picture elements:

* **1:1 Native Pixel Mapping:** If a display played video strictly pixel-for-pixel without scaling, a 1080p video (1920 × 1080) would occupy exactly 25 percent of a 4K display's physical surface area. The video would appear as a small rectangle floating in the center of the display.
* **The 4× Pixel Expansion:** To fill the 4K screen completely, the video must be expanded by a factor of two along both the horizontal and vertical axes (2 × 2 = 4× total area expansion).
* **The 5K Challenge:** On a 27-inch 5K Apple Studio Display (5120 × 2880 = 14,745,600 pixels), a 1080p video must expand by a factor of 7.1 times to fill the panel.

For every single source pixel in a 1080p frame, an upscaling system must synthesize three intermediate pixels to form a complete 4K frame.

Because computer screens cannot change their physical pixel grid to match incoming video files, the software or display hardware must synthesize intermediate pixels. The visual quality of the resulting picture depends entirely on the mathematical algorithm chosen to generate those missing pixels.

## How do traditional scaling filters work?

For decades, digital image processing has relied on classical mathematical interpolation filters. These algorithms do not "guess" what an object looks like; they evaluate the color and brightness values of neighboring pixels and calculate a mathematically smooth transition according to [libplacebo filter specifications](https://libplacebo.org/):

### 1. Nearest neighbor
The simplest scaling method. It looks at the nearest source pixel and duplicates it directly into adjacent spaces. While fast, it produces jagged, chunky square blocks. Nearest neighbor is ideal for 8-bit retro video games, but unwatchable for natural film.

### 2. Bilinear interpolation
Bilinear filtering samples a 2 × 2 grid of four adjacent pixels, calculating weighted linear averages horizontally and vertically. It is computationally lightweight and supported by every graphics card in hardware. However, linear blending softens high-contrast edges, making 1080p movies look distinctly blurry and smeared on high-density Mac Retina screens.

### 3. Bicubic interpolation
Bicubic filtering examines a 4 × 4 grid of 16 surrounding pixels using third-order polynomial splines. By considering a wider radius of pixels, it preserves sharper edge transitions than bilinear. Common bicubic implementations—such as Catmull-Rom and Mitchell-Netravali—balance edge sharpness against faint ringing halos.

### 4. Lanczos interpolation
Lanczos filtering uses a windowed sinc mathematical function, typically sampling a 6 × 6 grid of 36 pixels (Lanczos3). It preserves high-frequency spatial details up to the theoretical Nyquist limit of the source video. Lanczos yields crisp lines and sharp text, making it a favorite among cinephiles, though it can introduce subtle halo artifacts around high-contrast edges.

To see how these algorithms compare mathematically and visually, read our detailed guide on [video scaling filters explained](/maclc/blog/video-scaling-filters-explained/).

## What is AI upscaling, and how does it differ?

In recent years, machine learning has introduced neural super-resolution to video processing. Rather than applying a fixed polynomial formula across adjacent pixels, AI upscaling uses deep convolutional neural networks or generative models trained on millions of high-resolution images:

* **Pattern Recognition:** An AI model recognizes visual structures—such as human eyes, hair, foliage, brickwork, and asphalt—and predicts what those textures would look like at 4K resolution.
* **Detail Hallucination:** Unlike classical filters that merely smooth transitions, AI models synthesize entirely new fine detail. A low-resolution eyelash that appears as a single blurry dark line can be rendered as distinct hair strands.
* **The Artifact Risk:** Because AI models generate synthetic information, they can produce unnatural artifacts. Human faces can develop an artificial "plastic" or "waxy" sheen, straight architectural lines can wobble, and subtle natural film grain can be mistaken for digital noise and erased.

To understand the practical realities and processing costs of machine learning upscalers on Apple Silicon, see our breakdown of [AI video upscaling on Mac](/maclc/blog/ai-video-upscaling-mac/).

## What can upscaling never add?

The most widespread misconception about upscaling is that it can transform a 1080p video into true native 4K.

In physical optics and digital imaging, information theory dictates that once high-frequency spatial data is lost, it cannot be reconstructed with authentic accuracy:

1. **Lost Optical Detail:** If a camera lens captured a wide stadium shot where distant spectator faces occupied only 3 × 3 pixels in 1080p, no upscaler can recover the actual identity of those faces. Classical algorithms will blur them, and AI algorithms will invent generic faces that were never there.
2. **Compression Damage:** Consumer streaming video is heavily compressed using lossy codecs like H.264 or HEVC. During encoding, high-frequency textures, delicate shadow variations, and fine gradients are discarded to save bandwidth. Upscaling scales the compression artifacts alongside the picture.
3. **Color and Dynamic Range:** Spatial upscaling affects only pixel dimensions. It does not expand 8-bit color into 10-bit color, nor does it convert Standard Dynamic Range (SDR) into genuine High Dynamic Range (HDR) unless paired with dedicated inverse tone-mapping pipelines.

Native 4K camera capture contains 8.29 million pixels of authentic optical scene information. An upscaled 1080p video contains 2.07 million source pixels and 6.22 million interpolated pixels.

As explained in our comparison of [720p vs 1080p](/maclc/blog/720p-vs-1080p/), authentic visual clarity begins with the quality and bitrate of the source master, not the scaling filter applied afterward.

## Where does upscaling happen: Player, Mac, or TV?

When you watch a video on a Mac, upscaling can occur at three different stages in the pipeline:

### 1. Inside the video player
Modern media players utilize dedicated GPU shaders written in [Apple Metal](https://developer.apple.com/documentation/metal) to upscale video frames during rendering. This provides the highest quality because the player has direct access to the uncompressed, decoded YCbCr video buffer before operating system color management is applied.

### 2. In the macOS compositor (WindowServer)
If a video player simply renders a 1080p video surface and leaves display scaling to macOS, the operating system's Quartz and WindowServer compositors stretch the window using hardware bilinear or bicubic filtering. This uses minimal processing, but often softens the image noticeably.

### 3. In an external 4K television
When connecting your Mac to a living room TV over HDMI, you can output a native 1080p video signal and let the television's internal processor (such as Sony's Cognitive Processor or LG's Alpha series) perform the upscaling. High-end TV processors incorporate spatial databases and edge enhancers tailored specifically to large living room screens.

MacLC decodes in hardware through VideoToolbox, and its Metal renderer provides nine upscaling methods—from MetalFX spatial, AMD FSR 1, NIS and EWA Lanczos to neural passes for 8-bit SDR video (Apple Neural 4x, ArtCNN Anime, Anime4K, FSRCNNX). Its optional SDR to HDR engine also lifts highlights on capable displays while keeping faces and shadows close to how they were filmed.

## How to get the best upscaling results on your Mac

To maximize the visual quality of non-4K media on your Mac:

1. **Prioritize High-Bitrate Sources:** An uncompressed or high-bitrate 1080p Blu-ray rip will upscale to 4K far more cleanly than a starved 1080p web stream full of compression noise.
2. **Use a Player with Advanced Scalers:** Avoid players that rely solely on system-level bilinear stretching. Choose a player that offers customizable bicubic, Spline, or Lanczos filters.
3. **Avoid Aggressive Sharpening:** Over-sharpening upscaled video accentuates film grain, introduces glowing halos around characters, and creates visual fatigue.
4. **Match Your Output Resolution:** Ensure your Mac is set to output the display's native resolution so the monitor does not perform a secondary, redundant scaling pass.

Upscaling is an essential bridge that allows decades of standard definition and 1080p cinema to look crisp and clean on modern high-resolution displays. To learn more about moving up to Ultra HD, explore our complete analysis of [1080p vs 4K](/maclc/blog/1080p-vs-4k/).

## Questions people ask

### Does upscaling make a video look as good as native 4K?

No. While high-quality upscaling sharpens edges and reduces pixelation, it cannot recover authentic fine textures or high-frequency details captured by a native 4K camera sensor.

### Does upscaling improve video quality?

Upscaling improves presentation quality by smoothing jagged pixel staircases and preserving contrast on high-resolution screens, but it cannot add genuine visual information.

### Where does video upscaling happen on a Mac?

Upscaling can happen inside the video player using GPU shaders, in the macOS WindowServer compositor, or inside the image processing chip of an external 4K television.

### What is the difference between spatial upscaling and AI upscaling?

Spatial upscaling uses mathematical polynomial formulas like bicubic and Lanczos to blend adjacent pixels. AI upscaling uses trained deep neural networks to generate estimated details.

---
MacLC is a free, open-source video player for Apple Silicon Macs: https://www.hazenstudio.com/maclc/ · download: https://www.hazenstudio.com/maclc/donate/