Turning “Star Trek: The Next Generation” into HD cost CBS more than $12 million and took over three years. What you have is a laptop, a folder of aging camcorder files and a free weekend. That mismatch accounts for most of the disappointment people run into when they try to push video up to 4K.
Upscaling tools improve every year. What they still cannot do is invent detail that was never captured in the first place.
What 4K actually means before you convert anything
Terms like 4K and 1080p, or ultra-high and high definition, are really just descriptions of how many pixels are packed into a single video frame. Ultra-high definition — the thing nearly everyone means by 4K — measures 3,840 x 2,160 pixels. High definition 1080p comes in at 1,920 x 1,080.
Standard definition is rougher than memory suggests. Depending on your region it carries either 480 or 576 vertical lines, and the width varies according to how the footage was captured to begin with.
Cinematic 4K also exists, occasionally labelled “true 4K” or DCI 4K, and it sits at 4,096 x 2,160.
The arithmetic behind the leap from 1080p is what most people gloss over. Every pixel that was actually recorded has to be joined by three more on screen. That is four times the pixel count in total — and only one in every four is genuine.
Why upscaled footage looks soft
Any output file can be pushed to 4K. Frame size is nothing more than a number. Doing it without sacrificing quality is where things fall apart, and it is very nearly impossible, because altering the resolution does not reveal some sharper version of the file hiding underneath.
Conventional upscaling software bridges the gap with interpolation, enlarging each frame and figuring out how it ought to look. Everything hinges on how accurate that figuring is. The software has to estimate the values of the missing pixels, and genuine detail that was never recorded stays gone.
The result is usually softer than what you started with. All that guess-work used to manufacture the extra pixels ends up smoothing the picture over. Sharpening can rescue the edges, but it is a Goldilocks problem: apply too much or too little and the footage is wrecked. Noise reduction sets the same trap — push it too far and the image stops looking real.
The Star Trek restoration is the honest benchmark
Professionals working toward 4K begin with the finest source data available to them, which in practice means the original files. Better material going in produces a better result coming out. That is not a stylistic preference; it is the entire game.
The HD restoration of “Star Trek: The Next Generation” demonstrates the lengths required to do this properly. CBS ordered the upgrade in 2011 and returned to the series’ original 35mm film negatives — described as offering quality equivalent to a 20 megapixel resolution — in order to remaster all seven seasons from broadcast quality up to 1080p for Blu-ray.
The project drew on 25,000 reels of original film stock, with visual effects and CGI upgrades layered in along the way. It ran for more than three years and, according to one of its producers, reportedly cost the studio over $12 million.
And here is the irony. Moving analog film into the digital realm is the easier version of this task. Negatives retain far more detail than the television broadcasts of the era ever transmitted, so there is genuine information sitting there to be recovered.
Begin instead with an older digital master locked to a fixed resolution and the pixels you have are the pixels you get. Once your best available source is already low resolution, every absent pixel becomes a guess. In the best case that guess is persuasive. It is still not 100 percent true to form.
Interpolation versus AI, and what each one gets wrong
You almost certainly do not have $12 million lying around. So at home the actual decision comes down to traditional interpolation versus the newer AI-driven tools. Each generates the additional pixels a 4K frame demands. Their methods differ.
Interpolation is powered by old-fashioned mathematics. It selects a pixel within a single frame, examines the color and brightness of everything adjacent to it, and then computes what the surrounding pixels ought to look like. In its crudest form, that amounts to duplication. Stronger algorithms make an effort to hold edges together and keep lines from turning jagged.
Its output is predictable, and that is simultaneously its charm and its ceiling. Because interpolation cannot rebuild authentic detail or texture, the finished image comes across as faintly artificial. Your television may well be performing this trick in real time already, to inconsistent effect.
AI upscaling comes at the softness problem from another angle entirely. These models are trained on matched pairs of low and high-resolution images, which teaches them how familiar features are meant to appear. Instead of calculating pixel by pixel, the model forecasts what a sharper image would look like.
Sharper and more believable? Yes. A precise match for reality? No. AI upscalers also bring their own peculiar artifacts to the party, ones interpolation never generates. Since these models continue to advance, output that impresses today could look dated within a few years.
Old broadcasts bring an extra problem
Older video carries the added burden of interlacing, a technique that divided frames into two halves made of odd and even horizontal lines, each one refreshing as the screen refreshed. Contemporary displays do not operate this way. Vintage broadcasts designed for CRT-style screens did.
Those files have to be de-interlaced before any upscaling happens. Whether a 480i NTSC source can survive the trip to 4K and remain watchable is honestly still an open question.
Nearly every upscaling tool, AI or interpolation alike, hands you a generous set of settings to fiddle with. When the first pass misses, adjust the parameters and go again. Just confirm you have the disk space free, because one attempt will not be the end of it.
Resolution isn’t the only thing holding your file back
Do not feed a 1080p file into an upscaler and count on a clean 4K result. Technically you will get 4K, in that the frame will be larger. Arriving there without quality loss demands repeated attempts and a healthy tolerance for disappointment.
Small jumps beat large ones. Climbing from 1440p to 4K calls for far fewer invented pixels than 1080p to 4K, and dramatically fewer than 480p to 4K. Either way the software is estimating the gaps. AI makes that estimation simpler and possibly better, though a like-for-like match is never guaranteed.
Then there is bitrate, the factor nobody considers until it causes trouble. Compression shrinks files to conserve disk space, which means less data is retained for each second of footage. Going to 4K does not reverse that. The blocking and banding baked into a low bitrate source remain exactly where they were, just bigger.
Before installing a single piece of software, hunt down the best source file you own. If it is strong enough, everything that follows becomes considerably less painful.