March 24, 2022
One of the most common use cases I see for Video Enhance AI has to do with upscaling—or increasing the resolution of—videos. In fact, one of the most prevalent video question themes I get asked has to do with how to make videos HD or 4K. And there’s good reason for it! Depending on when a video was first digitally recorded or digitized from an analog source, the resulting clip could be very low resolution, making it look quite bad when viewing at full screen. And you probably know what I mean by “quite bad”. Have you ever viewed a low-resolution video at full screen? All of the details become blocky and it can almost be an impossible task to decipher what you’re looking at.
I’ve always been obsessed with everything related to outer space and space travel. Whether it’s following NASA’s developments around exploring other planets in our solar system or watching some classic Star Trek, I’ve always had a wide-eyed fascination with the universe outside of our own atmosphere. And while I’m so grateful to have near-instant access to archival footage of virtually every major NASA launch in history, I do lament over the low quality that so many of these clips suffer from.
Take a look at this still frame I grabbed from the NASA Atlas V launch containing the Mars rover payload, Curiosity. Do you see the blockiness and smudges throughout the entire still frame, especially in the smoke plumes and buildings surrounding the rocket?
That launch occurred on Nov 26, 2011... just over a decade from when this article had been written. While the prevalence of 4K or 8K video wasn’t as strong back then, the highest resolution public domain footage I could find from this historic launch is a lowly 720p, half the resolution of the more common 1080p HD resolution. This rocket’s payload consisted of Curiosity, a mechanical rover that was tasked with exploring the surface of Mars. You would think that such a monumental occasion would have warranted a higher resolution footage. Fortunately, advancements in AI-powered models makes it easy and attainable to convert low quality video to high quality. To start, take a look at this same still frame that was upscaled to 4K resolution, a 300% increase, using Video Enhance AI.
In addition to increasing the resolution from 1280 x 720 (720p) to 3840 x 2160 (4K), Video Enhance AI also did a wonderful job in adding important edge details, making every part of this clip more engaging and enjoyable to watch.
Before I share the steps I took to convert this low resolution video to 4K using Video Enhance AI, I’d like to share the source clip for reference. You can watch it below.
Now that you have a sense of what our source material looks like, here are the steps I took to determine which model in Video Enhance AI worked best for this purpose, which model settings were used, and which export settings I used to create the output file.
This first step is especially obvious, but I want to be thorough with my instructions. As such, the first thing I did was launch Video Enhance AI and click on the Browse button. I then navigated to the location of the source clip and opened it up.
Video Enhance AI comes with a nifty wizard of sorts that will suggest a model it thinks would work best on your video clip based on a series of options. To access the AI model picker, click on the light bulb icon towards the top of the toolbar. You’ll be presented with three variables to choose from, each with their own options. These options refer to certain characteristics of the source video, such as video quality, video type, and video artifact type. In addition to writing out the steps I took for the AI model picker, I also put together this video walking you through how to use it (as well as the the Comparison View, which I’ll go over in Step 3).
For this video, I selected Medium for video quality, Progressive for video type, and Blurry for video artifact type.
The selections you make will determine which model Video Enhance AI suggests. But remember: it’s only a suggestion! You are not required to use that model and, in fact, I recommend simultaneously comparing multiple models, which is my next step. But first, I want to highlight that the selections I made resulted in Video Enhance AI suggesting the Artemis Medium Quality model... a fine choice! BUT, there are several other equally capable models that I want to compare, so my next step is to click on the Compare button right above the recommended model.
Clicking the Compare button will bring you into the Comparison View mode. This is an extremely helpful view because it lets you simultaneously compare three AI models on the same clip. This is the best way to truly determine which AI model will best serve your needs.
Once in Comparison View, I click on the All Models icon, which is to the right of the Suggested light bulb icon. This brings up the list of all available models in Video Enhance AI. You’ll notice that the original clip will be parked in the top left quadrant, making it easy to compare. The top right quadrant has the suggested AI model, Artemis Medium Quality. Next, I’ll click on the lower left quadrant and choose Artemis High Quality, followed up with clicking on the lower right quadrant and choosing one of our newest AI models, Proteus. You’ll know which quadrant is active because the model name in the lower left corner of the cell will be highlighted in blue, while the inactive quadrants will have a translucent background.
Now that I’ve selected the three AI models to compare, I need to choose what my model and output settings will be.  I have several fine-tuning settings specifically with the Proteus model that I want to pay attention to. With that model’s quadrant highlighted, I click the Auto-detect Settings button (which looks like a lightning bolt). This will tell Video Enhance AI to analyze the current frame and choose the settings it thinks would be optimal.
The next setting I will choose is the output size. This setting will determine the upscale rate of the output video. As I mentioned at the beginning of this article, the resolution of this source clip is 1280 x 720, or 720p. I want to output it at 3840 x 2160, or 4K. That will yield a 300% increase in resolution while also improving image quality. But remember, 4K is not the only upscale option for HD resolution. With Video Enhance AI, you have access to a variety of common resolutions to meet whatever your video project needs are.
‍
Finally, I will choose the output video format options. Most of the time, I prefer the MP4 - H.264 option because it is possibly the most widely used format for sharing video online. Next, I sett he Constant Rate Factor to 0, which will yield the highest quality results.Step 5 - Previewing the AI Step 5 - Previewing the AI models
Now that all of the models and settings have been dialed in, it’s time to generate a preview. In the Preview mode, Video Enhance AI will process a predetermined number of frames from wherever the playhead indicator is on your loaded clip. In most cases, I’ve found that generating a 30 frame preview will give me enough information to choose the best model. So, the first thing I do is move the playhead indicator to an important part of the clip. In this case, it is a few seconds after the main engine ignites and the rocket begins to propel upward. Next, I enter in “30” in the # Frames field and then I’ll click on the blue Preview button.
Depending on the number of frames you want to preview, the hardware configuration of your computer, the quality of the source clip, and which models you chose to preview, it can take anywhere from a few seconds to a few minutes to generate the preview. When it is done, you’ll see all four quadrants play back in a loop, making it very easy to determine which model works best for your source clip. You can also pan and zoom around the clip in Preview mode.
Based on these results, I’m happiest with the Proteus model and that means that I’m ready to process my 4K output file. First, I will click on the blue “Close” button to exit out of the Preview Mode, returning me back to Comparison View. Finally, with the Proteus quadrant highlights, I’ll click on the blue “Start Processing” to begin the upscaling and enhancement on the entire video clip. Your computer configuration (including whether you are using a GPU) will play an important role in determining how long this processing will take. Other variables such as the original clip duration and upscale factor will also play a role in this.
However, once the processing is complete, you will be left with a higher-resolution video that has enhanced detail and improved quality. To illustrate what I mean, here is the 4K upscaled version of the Atlas V launch clip I shared at the beginning of this article. As you can see, in addition to the 4K resolution, the added detail throughout the video makes this amazing moment even more spectacular.
Now that you have a better understanding of how easy it is to upscale and improve the quality of digital videos, why not give it a shot with your own clips? Click here to learn more about Video Enhance AI and be sure to download a free trial today!
© 2005 - 2024 Topaz Labs. All rights reserved.
