Gigapixel v8.0.0

That’s the bittersweet part…

Here’s the dog’s friend (original online render):

Screenshot 2024-10-25 at 4.02.23 PM

Do some research into 3D (assuming you haven’t already) to see what’s recommended. The NVIDIA 4090 is very decent (and very expensive)… Maybe there should be a deal where NVIDIA gives out a copy of Gigapixel with each purchase!

The 4090 also comes in handy when using that NVIDIA Canvas app, as you have to have their hardware to use it. I suspect you can’t go wrong with it. Until the next one comes out-!

You’ll notice a bit of GP blocking or sectioning in the sky. Here’s another example:

But anyway, the detail is always incredible:

PS: You mentioned the speed of Adobe. It’s done online, not on-device.

Yes, before the cart one. A cart in front of a horse pushing the wagon down a hill. But maybe the horse is backing out? That leaves the AI ​​to answer for us viewers. But your other pictures are also beautiful. Including a cute dog. It’s amazing what you can do with graphics tools. As a new branch, it’s growing incredibly fast. Let’s keep our fingers crossed for the developers of new software tools.

:wink: :wink: :wink: :desktop_computer:

Without question, the 4080 SUPER and it’s future proof.

That said, NVIDIA will be releasing the RTX 50 series’ cards ~1Q25 (definitely future proof) but I’ll leave you to those decisions.

A future-proof GPU? Especially from a company that keeps pushing features that are only available on newer models? Good one.

1 Like

At least it isn’t LOAM.

Thanks! I am having a great time with this but am no AI pro (catching up). I started out with a Kodak Instamatic in the mid-'70s and now here we are! :wink:

Here’s a long creative story…

One of my first NVIDIA Canvas renders, June 2021. I drew some triangle mountains and other scenic elements:

e10

Redefined in GP on the PC earlier (reduced result here):

Image-to-image reinterpretation of the GP results in DiffusionBee with about 40% input strength, back to 512x512px output:

DarkFantasyArthilltopmoun_19804981

Attempting to get a Redefine preview of this new result at home on my M2 Mac Mini, oooh…

Forget it, I’ll just do a 2X Cloud render (1 credit, took about 10 seconds, but oops, I lost the 2nd castle):

Now to bring the 2X Cloud render back into GP on the Mini and do additional normal 4X upscaling, which of course cleans up the render without the artistic reinterpretation of Redefine (this is “poor man’s Redefine”!). It processed quickly:

I marked some additional DiffusionBee favorites (which I banged out quickly this evening on the Mini at various input strengths and models) to Redefine at 4X or 6X on the 4090 PC next week!

2 Likes

I’ve been wondering a definitive answer for a while Is Gigapixel superior to Topaz Photo AI at upscaling/restoration or vice versa or are they truly using the same models?

People say since Giga is dedicated for simply upscaling and so its better

ISSUE: dropped a few input images, for all enabled Redefine & Face AI.
In actual output Face AI is applied only to some images.
All input images actually have clearly visible faces.

image

a company that keeps pushing features that are only available on newer models

This has nothing to do with NVIDIA.

For GP 8.0+, TVAi 3.x.x to 4.x… Topaz, has nothing in the near future (next 2- to 3-years) that could max out the RTX 40 series’ cards (from the 4070 SUPER to the 4090) with a PROPER CPU, ect.

Very happy with V8 performance, but unhappy with this Cloud Render button that takes up half the space where Export Image is. I would appreciate an option to hide the button. I don’t need it and don’t want to accidentally click on it.

2 Likes

Yes, the Adobe Gen Ai is proverbial Cloud.

But it works! And, in about 20-30 seconds. I’m not getting results even after 20 minutes of GAI Cloud…

I’ve had no trouble in GAI cloud so far. Post a pic right now if you’re here and I’ll 2X it right away.

1 Like

You may refer to the following post that I posted before. I hope it helps.

1 Like

I began to observe very strange artifacts. Win 11


settings: creativity 1 / advanced settings: all 1

1 Like

ISSUE: GAI 8.0 is trying to making hair out of most patterns.
e.g. this was drapery in the input image

it is pretty incredible in itself, but this must be controllable, so it could retain drapery in the output

5 Likes

I’m having the same issue on a Dell laptop with Windows 10

The boat from Pt Reyes in the beta forum yesterday. It’s a different boat (it’s real!) than the gen ai on a dry ocean bed one I concocted previously.

I’m just waking up. I posted a zoomed in view there.

I also included a snip of what I wanted added in that other thread’s description box in my rpt there.

Will post snip of full image here (it is soft and has some text on the boat) after I eat breakfast.

Thx!

Models of this v8 are very good, but again color spaces and bit depth are broken during export.

  • Option “keep color space” doesn’t work, I have to specify it explicitly.
  • Problem, after export, it is not written in the profile, so we don’t know the export kept it.
  • The program doesn’t work in 32 bit floating point (too bad for hdr images), fine. But again we have to specify it explicitly during export.

On a side note, the recovery/redefine modes need GPU to be used, OK, but I’d like to have the option to use CPU/GPU/GPU2… in the export panel, because for other rendering I use CPU, the results continue to be better.

The new model are impressive, but please take care of color profiles and bit depth, I use the program again, when this bug will be fixed.

Have you tried launching GAI as administrator (right-click the icon and chose “run as administrator”)?