The Truth About AI Video Enhancers?

The adverts sound great . . . “Turn your blurred and grainy standard definition video into crisp 1080p or even 4K”.

So what is the truth?

You must have seen it at the cinema or on TV numerous times . . . the detective gets a bit of dark, blurred CCTV footage and blows it up so crisply that you can read the car number plate at 50 metres in a dimly lit street or see the tattoo on the wrist of the mugger.

Well it ain’t like that!

For my tests I used the HitPaw Video Enhancer AI

They do work to improve the image but don’t raise your expectations too high as it all depends on your footage.

The best results appear to come from slightly soft video such as decent quality SD DV or HD that hasn’t been focused too precisely.

If you have visions of restoring some appalling quality VHS from the 1980s . . . forget it unless they are close ups with not too much detail.

As you can see from these framegrabs of a standard definition DV clip, there is a worthwhile improvement in cleaning up the image and adding a bit of sharpening but this is about the best you are likely to get.

What isn’t talked about at all in the adverts is the processing time.

The example I have given is from a 4 minute 38 second film shot on DV PAL (768x576) and to clean it up would take an estimated 2 hours 33 minutes.

Upgrading it to 1080p or 4K would take many times more and there would be no noticeable improvement over the standard def version.

Some users have reported estimates for an hour long video of over 100 hours!

My tests were done on a 2017 27” iMac which isn't fast but doesn’t exactly crawl. Obviously the latest machines could dramatically reduce the processing time but it would still be pretty lengthy.

So my conclusion is that video enhancers are fine on powerful machines, treating reasonable quality video that is quite short in length!