Topaz Gigapixel AI 6 – Time to Face the Truth?
Readers of this blog will know I am a big fan of Topaz Gigapixel AI. I have owned it since version 1 and I would not enlarge a photo without it.
Recently, I received a phone call from someone who wanted to make a mural of one of my photos. A TWELVE FOOT WIDE mural. I enlarged the original to six times its original size using Gigapixel. My customer was delighted with it once it was up on the wall and sent me this photo:
But there’s one area where enlargements are really difficult – those containing human faces. I don’t know how Topaz implemented Gigapixel AI, but to do that kind of enlargement, it has to add detail to the enlargement and also rely on the human brain to ‘see’ the result despite any imperfections introduced along the way. Our minds do that pretty well – a tree, a mountain, a field, the sea – all these scenes have a randomness to them. Even if Gigapixel AI makes a mistake, as long as it looks like a tree, a mountain, a field, the sea then our brain accepts it as it still looks realistic.
But human faces are much harder to enlarge, because we scrutinize them more, we know when something looks amiss. It sticks out like a sore thumb, so to speak.
Early versions of Gigapixel AI received some criticism when used on pictures containing faces. Some, of course, were ridiculous – people expecting an image the size of a postage stamp to to enlarge up to a 30 inch print. But some were justified – the faces just didn’t look realistic enough.
At the time, I didn’t care too much. Most of my photos had no people in them. But then I shot a couple of weddings and suddenly enlarging pictures with people in them was more important to me.
Topaz continued to enhance Gigapixel AI and now, from version 6, it has a mode especially designed to work with human faces. But is it any good?
Gigapixel AI vs Affinity Photo Face Off
For my test photo I chose a wedding picture featuring the bride and groom. I developed it in DXO Photo Lab and used Portrait Pro to get the results I wanted. Then I *reduced* its size to a 2000 pixel wide image and saved it as a jpeg. This is a reasonable test scenario. It’s effectively what many people may have got with their mobile phones, or perhaps a very tight crop of a regular picture.
Gigapixel AI vs Affinity Photo – The Test
The test was to take my 2000 pixel picture and enlarge it so that a 12 inch and a 30 inch print could be made. That is enlarging it to a 3,600 pixel wide image and a 9,000 pixel wide image.
For the purpose of comparison, I used Affinity Photo to enlarge the same picture to these dimensions, using the Lanczos algorithm. Affinity Photo is my favourite general purpose photo editor. It’s a great product.
The Test Photo
Topaz Gigapixel AI vs Affinity Photo – 12 inch enlargement
I wasn’t surprised at the result from Affinity Photo. But I was astounded at how well Topaz Gigapixel AI did. Just look at the comparison photos:
Affinity Photo on the RIGHT, Gigapixel AI on the LEFT
In each case, Topaz Gigapixel AI has enlarged the photos without introducing blemishes and artefacts.
Topaz Gigapixel AI vs Affinity Photo – 30 inch enlargement
The Affinity Photo results are essentially unusable. Let’s face it, I’m doing an almost 4.5 enlargement of a 2000 pixel jpeg. I expected rubbish and got rubbish.
But just look at Topaz Gigapixel AI’s result. I’ve applied no sharpening or any other tweaking to it – this is the result straight from Gigapixel. Could I make a high quality print from this image? Easily!
This is the best result I’ve got from Topaz Gigapixel AI. It’s an incredible result.
Affinity Photo on the LEFT, Gigapixel AI on the RIGHT