Review on YouTube: Best A.I. Image Upscaler? Top 7 Software Compared!

PiXimperfect (1,62 Mio Subscribers) published today a review of seven Image Upscaler programs on YouTube:

Best A.I. Image Upscaler? Top 7 Software Compared! (PIXImperfect on YouTube)

The result unfortunately did not go so well for TopazLabs GigaPixel AI, the relevant passage starts from minute 15:05 in the review.
My trial version has unfortunately expired, so I can’t reproduce the test result and comment on YouTube if this is a bug or problem on the side of the reviewer.

Sorry for the bad news.
Best Regards and a happy holiday season!
Juergen

3 Likes

I think the artifact issue has been raised before.
PITY!

1 Like

I suspect Dinda’s computer did not have the latest drivers. Those lines are more hardware-based than that made by Gigapixel. He seems to forget that there is too an online version of Gigapixel … but I must admit, the eventual winner was impressive, if only for 19 free images a month!

2 Likes

Curious, which app came out on top?

1 Like

Hello,
imglarger.com” (a web based service) was first and “letsenhance.io” second.
Especially letsenhance.io is very difficult to comprehend, because it removed the lower eyelashes in the sample picture, an absolute no go in my opinion. Why upsample a photo if you lose detail in the process? As a curiosity: the best Photoshop method came in last place, embarrassing for such a big and rich company.

I’m sure GigaPixel AI would have had a big chance to win without this nasty bug(?).

Regards Juergen

1 Like

It’s funny. I triad this side and I put one picture it was a little bigger so rejected. I put second picture and it tell me that I used up the amount of pictures and it start forcing me to sign up. SCREW them…

He should have used CPU processing instead of GPU processing. I think the GPU processing on some machines messes up the result. I also realised that some images won’t be improved anything much for an still unknown reason. In this case reducing the image size before enlarging might help. I could imagine that it has something to do with the detail density of the image. If you have a 1080p image out of an action camera where the camera took a 720p cut for stabilising and when the camera enlarged the image to fake 1080p again, the detail density of the image does not match the pixel density anymore. This may cause the software to struggle with.

Apparently you sign-up - but if you stay under 20 images a month - no payment is taken.