[Fenwoodian]

.
A couple of days ago I purchased Topaz's new software Gigapixel AI. I've been running a few tests and so far I'm quite impressed.

It up-sizes images. But more than that, it also reduces noise in the image's out of focus areas, and it not only up-sizes but it uses AI to sharpen the image as well.

I've used many types of software (too many to mention) over the years for noise reduction and for sharpening. This Gigapixel AI does a better job than any of them.

In fact, the results are so good, that I deleted most of the other noise reduction and sharpening softwares that I had on my desktop computer.

You need a modern/fast computer for this software.

You Tube

You Tube

I've currently got a couple of 16MP APS-C cameras that I was thinking of upgrading to 24/26MP newer bodies. But after seeing the results I'm getting from enlarging tiny old/ancient image files AND files from my modern cameras, I believe that I'll be keeping my 16MP bodies because Gigapixel AI will enlarge and denoise/sharpen these 16MP files to as large as I'll ever need for my prints.

---

Last edited by Fenwoodian; 10-04-2019 at 03:39 PM.

[Wingincamera]

I downloaded the program about a month ago to try out. I wasn't that impressed as far as the sharpening. It didn't look any better that what I got in Lightroom. I uninstalled it.

[Fenwoodian]

I downloaded the program about a month ago to try out. I wasn't that impressed as far as the sharpening. It didn't look any better that what I got in Lightroom. I uninstalled it.

.

How large did you enlarge your images (2X, 4X, 6X)? Did you do actual side-by-side print comparisions between Lightroom and Gigapixel AI? If so, what size were your test prints?

[vonBaloney]

Gigapixel AI is more for enlarging that sharpening. They have another on called Sharpen AI for sharpening (which does more than normal sharpening -- it rescues missed focus and fixes camera movement) and Denoise AI for noise. They are all pretty amazing, and they update the models they use quite often. You do need a decent graphics card.

[clickclick]

I like their DeNoise too. I haven't tried Gigabpixel AI, but I'm not surprised you're getting good results based on my experience with their products - also use Topaz Studio and Clarity and others.

[Fenwoodian]

.
I have a large format printer and take photos for the primary purpose of making large prints. As such, this is important software to me. But for those who just post photos on the Internet, I would not see a lot of value in this software.

The other use for Topaz Gigapixel AI is for salvaging old digital images, or for cleaning up and enlarging scanned 35mm negatives and slides.

[Fenwoodian]

Gigapixel AI is more for enlarging that sharpening. They have another on called Sharpen AI for sharpening (which does more than normal sharpening -- it rescues missed focus and fixes camera movement) and Denoise AI for noise. They are all pretty amazing, and they update the models they use quite often. You do need a decent graphics card.

.
I'm convinced that there's some of the "Sharpen AI" DNA in "Gigapixel AI". Every image I've used in Gigapixel has had a healthy dose of sharpening applied to it. In fact, I can't imagine use any sharpening software (including "Sharpen AI") on an image that I'm eventually going to use Gigapixel AI on.

---

Last edited by Fenwoodian; 10-05-2019 at 02:16 PM.

[biz-engineer]

The market of pixel peeping is still going strong.

[texandrews]

I trialed and then purchased Gigapixel AI for a professional project with very demanding specs. I can't show you the images, unfortunately due to contract restrictions. But I can tell you what I was asked to do and did.

The Hirshhorn Museum in Washington DC will be mounting an exhibition of a gift of Marcel Duchamp's ephemera and smaller objects/multiples. Two of these are iterations of "The Green Box" and "The White Box". These boxes contain assembled notes for his work. The designers and curator of the exhibition wished to have a way to display the contents during the exhibition, but this requires constant rotations as the materials cannot be exposed to even the low FC levels of the museum. They cooked up the idea to have a 1:1 image of the contents as a photo background on which individual bits could be rotated and laid on top in proper position, preserving an exact 1:1 correlation between image and artifacts. The curator assisted by a conservator and the designer arranged all of this on a sheet of foam core, and the whole bit was wheeled into the photo lab and shot on a large motorized copy stand using high cri color adjustable LED lights, but no polarization. Shot with a 645Z. The array was too wide for my 645 55 lens at full column extension, so I had to use the DFA 35. Something of a torture test in more ways than one. If you ask why we didn't stitch, there were multiple technical/logistical reasons during the shot, and I wondered in this case if the stitch wouldn't bring in its own problems (pretty densely detailed image, and due to the nature of the materials pretty easy to see any stitching goofs...)

The whole desired image of each exceeded the 1:1 capabilities of of even my 645Z at 300ppi, so some uprezzing was required. I uprezzed one 1.5x and the other 2x. The design team decided to convert the photos to B+W so that the originals could be very easily distinguished against this photo background by the public. All of this will be displayed flat in plexiglass topped vitrines, the overall size about 44-ish inches on the long side.

Yesterday we finally pulled a full size test print off of the graphics department's plotter-printer (so not the best print for photo). For the Green Box materials, arrayed in a dense overlapping cluster, I was quite pleased with the results resolution-wise----a very clean result. it was better than I expected based on the image on the screen. The other image has not been printed yet, but it looks better on the screen than this first one, so I expect it will be quite good.

That said, I had my doubts as I was processing the images. In initial processing trials from tiffs out of LR, Gigapixel AI's auto settings produced alarming halos and strange "vermiculation" artifacts among areas of random textures. I had to dial down my normal sharpening and contrast settings in LR quite a bit and reduce output sharpening to low, and I had to use the manual settings in Gigapixel AI and reduce/tweak the default setting there as well. There are only 2 manual slider settings unfortunately, IIRC.

I hope to have time to produce a Youtube video on this when I have time(cue hysterical background laughter) in the next few weeks.

So, to recap the salient points and add a couple:

• Good to excellent results are possible when used with care. Combined with judicious leverage of dpi in a print very large prints can be produced.
• Care is absolutely required!
• The program adds sharpening and contrast, so pre-process accordingly!
• It's an open question how far you can uprez---I bet it's very image dependent.
• The software needs more and better manual controls
• The "AI" part needs further development
• Odd artifacts can be induced---watch those blank or random textured areas
• I bought the software at full price---I buy things I need but I don't spend frivolously, and my money is a vote of approval.

I am anxious to do some trials of this with some of my own artwork images. For me, a 40-60-ish inch long side dimension for the things I am doing right now, multiple vertically oriented shots in n-tychs would be a sweet spot. With a triptych, for instance, I could have an overall framed dimension of floated images about 64-ish by 140-ish inches, which is as big a work as I want to handle (and pay for, even printing and framing/glazing myself) right now.

[Fenwoodian]

I I had to use the manual settings in Gigapixel AI and reduce/tweak the default setting there as well. There are only 2 manual slider settings unfortunately, IIRC.

.
Tex, thanks so much for sharing! You certainly have much more experience with this new software than I do.

I too always use the "manual settings", and like you wish that there were even more manual settings.

For many of my images, I actually increase the suggested manual settings. I exclusively deal with images with most areas being out of focus (background blur) and only a smallish percentage of the canvas being in focus (the subject). I like how cranking the two manual settings to their max seems to be optimum for enhancing the bokeh of my images (I have done many hours of side-by-side comparisons of the effects of the many variations of various manual settings have on my bokeh photographs).

Conversely, I can detect no visible differences in having the "face refinement" toggle on or off for my images that have no faces in them. So, I just leave the face refinement switch always in the "on" position.
.

[clickclick]

The program adds sharpening and contrast, so pre-process accordingly!

For their Denoise program, they recommend using it before post processing if possible or minimal processing. I'm wondering if this applies here too. I found this video on prepping your RAW file in LR helpful when I started using Denoise 6.

You Tube

You Tube

Sounds like a great project to be involved in, and really nice it's being done with Pentax equipment.

[manishved]

For their Denoise program, they recommend using it before post processing if possible or minimal processing. I'm wondering if this applies here too. I found this video on prepping your RAW file in LR helpful when I started using Denoise 6. Topaz DeNoise 6 - How to Be a Noise Reduction Ninja - YouTube

Sounds like a great project to be involved in, and really nice it's being done with Pentax equipment.

Watched the video and I will give it a try. Thanks all!

[interested_observer]

I've been somewhat considering their AI Denoise product for awhile, primarily for the landscape segment of astrolandscapes. With the small amount of light you are capturing in absolute darkness, you are going to get noise. It's just the fact of life. Having said that, I've found a way around the white dots of the K1 (Raw Therapee - impulse noise slider) and the rest of the noise can be substantially reduced, however it does appear that their denoise product potentially does a better job, without the artifacts after effects.

• Has anyone else used their denoise product on K1 images captured in the pitch dark?

• Also, does anyone know how the product handles the white dots from the K1?

I really need to spend an entire weekend moving images to my main system - cleaning up things, combining catalogues, and then ordering a new graphics card that denoise requires. I've just been so lazy. I also need to order a new monitor that will do a better job of displaying the images. Just gotta get my s**t together.

[Fenwoodian]

.
interested_observer, you can download a free 30-day trial of Topaz DeNoise AI here. Feel free to try it out at no cost.

[vonBaloney]

Please note the Denoise AI is not like their old Denoise program (which is also excellent). And you probably should eliminate any other denoising/sharpening going on in LR or wherever before you apply these. I'm not positive if Gigapixel should be a first or last step, but I'm assuming last, otherwise any additional processing will be very computationally expensive because the image will be that much bigger. I have mainly personally used Denoise AI and Sharpen AI -- my wife is the "preparing photos for printing" guru, I'd have to ask her about that. I do know she said Gigapixel works great with some images, some others turn out weird. Which is to be expected with this tech. (Just as Sharpen AI can rescue some images amazingly from camera movement or slight missed focus but others will have too many weird artifacts.) All worth having in the toolbox though.