Introducing AI Gigapixel with Greg Rostami - Updated Presentation (v2)

What will you do with 3600% MORE PIXELS?!

This updated webinar recording showcases several examples of how GigaPixel can “A.I. Imagine” all of the missing details when enlarging photos.

Making the perfect crop … for example, wildlife photographers often must frame their subjects from further away to capture the perfect moment. Convert your DSLR / Mirrorless photos into Gigapixel images that can be printed on walls that are 10 feet tall at 300 DPI. Enlarge smartphone photos to unparalleled resolutions for a variety of applications. Capturing high-quality stills from video to capture that “perfect moment”.

One of the most exciting features of GigaPixel is how it recovers incredible detail from highly compressed and noisy photos. The webinar will also show a typical workflow for enlarging a small crop from a photo and then comparing that to other interpolation methods.

A Topaz Live Training Recorded: September 4th, 2018

[natureupclose]

Will this software be used in video?, i have tons of SD content that can use an improvement.

[kuyajaypi]

How is A.I. Gigapixel different from Sharpen A.I.? Doesn't Gigapixel just sharpen the resized image?

[bobsagget]

Mistake #2:
19:11 You compared photoshop vs original when you claimed you were comparing photoshop vs your neural net

[yodawok]

Good presentation.. a little too much of samples (Unfortunately I began to loose interest) ; however my question (before I buy) is there a workflow?

[myhaverphotography]

I have to say that I am impressed with AI Gigapixel post the update. This is not based on your presentation samples but my own side by side comparisons. Quite frankly I get much better results from Adobe Photoshop “Preserve Details 2” than what you have presented.

Prior to the update I did several comparisons and my results proved AI Gigapixel was not ready and very unimpressive. I actually uninstalled it. A few days later when I heard about the update I thought I should give it another try. Now I am very impressed based on my own comparison results. Great improvement and nice work, Topaz!

BTW, I had the same initial disappointment when testing AI Clear. It did not impress me in the least and so I did not purchase it. Has that also been improved?

I will be appraising my needs for upscaling Images, not something I often find necessary, as to assessing my justification of the expense of the AI Gigapixel software.

[dbrown]

you need to link to your website in the description

[Randyrocker]

All I have is version 1.1.1, so where is Version 2 coming from?

[berniejii]

I could not find the blog post by the CEO. Please give a link.

[hypersonicmonkeybrains]

Trial ended and i purchased. it's just too good i can't go without it!

[bloodywolftr]

Ok i have tested version 2.1. So the app requires some kind of good image sample even on small sized pictures.most of the time the algorithm work good on everything other than human beings in a little bit of distance.It is not working smart as presented in the video.Why? Cos it has no differentiation between body parts and the face.Especially on the eyes of human it messes up.Irises gots bigger goes outside the eyeballs, collide with eyelashes etc. The model needs to be trained on people faces in a distance.The nature photography it is good.Also worth to mention not giving good results in some scenarios of noisy pictures.But i have to say it has a very prommising future as a product.Keep up the good work topaz team. Note : What i suggest or request is implement a feature that gigapixel can detect faces of human, which is available with tensorflow, i have seen examples on github if i am not mistaken, and with that in consideration, gigapixel works on mixed mode, face region with face algorithm(which should be trained for it) and the general algorithm should work on the rest of the picture.Also a good option should be let users select the areas with faces if the detection fails, user input style.

[bgtubber]

Awesome stuff! A little correction at 39:28 - 4 times faster, not 400 times. 🙂

[hypersonicmonkeybrains]

why does it not have Ai downscaling as well? since most people don't want huge size images. if it had both, it would be a killer app.

[michaels]

Thank you, but I agree with another comment that many fewer words (and skipping some examples) would have been better.

[traveler]

No thanks, just gave it a 30 day workout. Stalled/Crashed at least 33% of the time.

[jimcochrane]

It does strike me a bit too long, and bit repetitive.