HitPaw VikPea Video Upscaler Review

The past couple of years have seen the emergence of AI-based video upscaling software hitting the market and HitPaw’s VikPea Video Upscaler has entered into that space as a good example.

I have never bothered with any video upscalers or upscaling features in editing software until now because basically all of them have been pretty much rubbish.

You can read all about why that has been the case prior to the introduction of A.I. into the picture in this more detailed article on video upscaling here.

If that’s all a bit wordy for you the bottom line is that just doubling the number of pixels in an image is definitely not going to give you a better image!

So the big question is whether HitPaw’s video upscaler, VikPea can actually deliver on that promise of higher resolutions while improving or at least maintaining image quality.

Interface

First up here’s an image of the interface which as you can see is clean and remarkably simple to navigate given the complexity of the task at hand.

Anyone who has wrestled with some of the more high end upscaling solutions out there for professional use will appreciate this approach.

At the most basic level all you really have to do is decide what kind of footage you have based on the examples given on that screen.

So in my case I choose the DVD Videos option you can see on that screen above.

I chose that one because the footage I wanted to try was from an old Sony DV Cam shooting at 720×576 Interlaced at 25fps which was the DVD standard for AVI files to be burned to DVD at that time.

So for this test I was asking the program to not only convert to 4k footage but also de-interlace that footage to retain smooth motion!

VikPea automatically directed me to the General Denoise model and I just let it create the preview you can see below.

Although it may be hard to tell from that image the result was exceptionally good.

The smooth motion of the video was retained, there were no interlacing artifacts and overall the image has been sharpened in a quite natural way.

The most important point being that although the video is now at 4K resolution there has been no loss of definition and in fact the image quality overall has improved.

Each choice on that opening page is optimized for different types of footage and the results can vary depending on which one you choose.

Regardless of which model you go for, you can have the software output a small sample of the upscaled footage so you can decide whether you are good to go or want to try another one.

The HitPaw VikPea AI models.

The General Denoise model is designed for footage that has some noise in it and this very often comes from lower quality or resolution footage shot in low light, less than ideal conditions or older equipment.

To be honest this one is probably going to be the one most people use as it give great overall results under most circumstances.

The Animation model is specifically designed for cartoon content and does a remarkably good job of maintaining those clean lines and solid color areas without introducing the weird artifacts you often see in other upscalers.

As for the Face model… this is where things get interesting.

It’s clearly been trained to enhance facial features and does an impressive job of maintaining skin texture while sharpening essential details.

However, you wouldn’t want to use this for general footage as it can create some pretty bizarre effects on non-facial areas.

On the footage I tested I assumed that the “Sharpen” Model would give better results on low res footage and often this was the case.

However just as often similar low res footage with faces actually looked better after using the “Soften” model so it’s well worth testing those ones in the preview before proceeding.

The software supports upscaling from 480p all the way to 8K, but let’s be realistic here… pushing footage from 480p to 8K is asking for trouble!

The sweet spot seems to be going up by 2x or maybe 3x your original resolution.

Processing Speed

Processing speed is where things get a bit technical and we have to start getting a bit realistic about expectations!

The software leverages GPU acceleration and that certainly makes all the difference.

The reality is that VikPea has to analyze each individual frame as it is moving through the footage and then apply A.I. learning to that frame before moving onto the next.

However because it also has to be reading earlier and later frames as references for both content and motion plus reconstructing the video file on top of that… well it all gets a little intense processing wise!

I ran my tests on my computer which has an older Geforce GTX 1050 Ti.

On this setup I was getting speeds of about 1 second per frame of upscaling on the demanding test I described above using the old Sony DV tape footage.

On a different computer without a dedicated GPU using CPU only, things were about 4 times slower and to be honest I gave up on it all together.

My advice would be to test it on your system using the trial version before committing to purchase or any big projects.

Different AI models and scaling factors can result in processing times varying enormously depending on your configuration.

Intelligent Artificial Intelligence?

Just for a bit of visual reference here’s a video covering some of the features in HitPaw VikPea.

One particularly strong feature of the program is the use of the suggestions on the Home screen as shown above to let the software choose the correct model for the footage you have.

The reason this works so well is that in developing VikPea their engineers would have had to have identify common problems in videos from a wide range of sources.

Over time they would have built up a database of types of footage and the most likely solutions to those problems.

In my testing the suggested model for the footage I was dealing with turned out to be the best one in every case.

The Bottom line on the HitPaw VikPea Video Upscaler:

It’s a solid piece of software that does what it says on the tin.

The interface is intuitive, the results are generally impressive (especially with the right AI model for your content), and the performance is strong if you’ve got the hardware to support it.

Is it perfect? Nope!

The “Face model” can sometimes be a bit weird and you’ll need some decent hardware to get the best out of it.

But for anyone needing to upscale footage regularly, particularly if you work with different types of content, it’s definitely worth considering.

Just remember to test it thoroughly with your specific hardware setup and the type of footage you typically work with.

See HitPaw VikPea Here for Yourself