Gigapixel v7.2.0


We are excited to announce a big update today with Gigapixel v7.2.0!

You can download this version of Gigapixel using the links below:

Released June 5th, 2024
Mac: Download
Windows: Download

This is another important milestone for Gigapixel, this release includes:

  • New Speed mode for Recovery that is 20x+ faster, with lower system requirements. No mega-GPU needed - this should run on most machines that can currently run our other upscaling models.

  • New Multi-Preview for Recovery, allowing you to generate multiple preview patches before committing to rendering the entire image. It’s also frameless, and it just looks phenomenal compared to the pop-up window that blocked your workspace before.

  • You can now switch models inside the Compare Tool without having to leave and re-enter that workflow, making the process more seamless and useful if you’re in a hurry.

  • You can also eliminate models inside the Compare Tool to narrow down and find the best results faster than ever before. There’s also a new unique split-view with 3 models.

  • You can now import CMYK files and export CMYK TIFF files as well. As a result you should no longer have to convert files for direct printing before and after using Gigapixel.

  • Model updates with improved Low Resolution v2 results. Less artifacts, more sharpness, and better results across the board using this new model.

  • Under the hood we have also improved color accuracy inside Gigapixel. You should now get more reliable results, with less issues using high-fidelity colorspaces like Adobe RGB or Pro Photo.

We put together a quick video overview as well to help you get the most out of these new features:

And here’s a full breakdown of what to look for in this latest release:

Recovery Model Multi-Preview

This version of Gigapixel also brings a major improvement to the way previews are generated for the Recovery model with a new borderless preview inside the workspace.

You can still press the spacebar to show the original image underneath to compare.

Also, you don’t have to stop at one patch! Check it out:

You can now preview multiple different areas of the image by generating new patches. Each patch will be linked to the same “detail” slider, so you can understand how the detail changes will apply to important areas of the image.

10x Faster Recovery Mode

The results we’re seeing from the community from the Recovery model are absolutely stunning, but we are aware that certainly not everyone has access to the sheer GPU firepower required in order to run them. So our team has been working hard to make these models both faster and less demanding so that more people can access them.

Now, with the latest release of Gigapixel, you can use a new “Speed” mode for the Recovery model, which is dramatically faster than the prior version of Recovery, which you can now access with the “Quality” mode. Check it out:

Quality maintains the amazing output results from Recovery you have seen in previous versions. This will still be the best option as long as you have a powerful GPU.

Speed is a new faster version of our Recovery model, allowing for quicker generations for a small quality tradeoff. This mode also produces fewer artifacts than the Quality mode.

The requirements for the Fast modes is much lower, and should run on most machines if you are able to run our other GAN models. They will still be slower than our other models, but internal benchmarking shows significant speed improvements over the current Recovery model:

Recovery (Quality): 1x speed, current model
Recovery (Speed): +20x speed on average

As a result, you can also now push the boundaries with the Recovery model on larger images using these settings. You may also get better results at slightly larger image sizes with Recovery Fast than with the Quality model.

Above, this 640 x 853 px image ran in about 30 seconds on a Macbook Pro M3

In the example above, the same render would have taken upwards of 10 minutes on the same machine with the Recovery Quality model.

Compare Tool Expansion

We have improved the Compare Tool to help you find the best upscaling result faster than ever before. Now you can remove models from the comparison view, allowing you to narrow down your selections from up to 4 models down to a single one.


You also no longer have to exit Compare mode in order to pick a new model. When inside comparison view, you can now select a new model for each pane, swapping models until you feel confident in the results and in your model selection.

Recovery is not currently available within the Compare Tool

CMYK in. CMYK out.

You can now both import and export images in the CMYK colorspace for printing by selecting from a new colorspace dropdown menu in the export settings.

We have also made even more improvements for color accuracy and color management in Gigapixel under the hood as a part of this release.

Note: CMYK inputs from plugin workflows such as PS Automate are not currently supported.

If you select TIFF as an output, you will now see an option to export in U.S. Web Coated (SWOP) v2, among other color spaces:

Keep in mind that the preserve option also preserves the input colorspace.

We will be providing more options for output colorspaces and custom profiles as well, so please let us know what other colorspace options you’d like to see in Gigapixel.

Also, please note that PNG files only support the sRGB colorspace in Gigapixel.

Low Resolution v2 Model Update

You should now see even better results with the Low Res v2 model across the board.

We have updated the Low Res v2 model to improve performance and to resolve some artifact issues which appeared in certain types of photos - particularly those with high amounts of noise or grain.

With updated results bringing the Low Res v2 model out of Beta, this model is now included in auto model selection mode.

If you’re still a fan of the old model, no worries - you can still access the Low Res v1 model in the model dropdown. Just click the menu button next to Auto mode:

Screenshot 2024-06-05 at 1.13.20 PM

Screenshot 2024-06-05 at 1.12.50 PM

With your help, Gigapixel is getting better every single day. If you’re not already in our Beta group, you should be! If you have an active Gigapixel license, you are welcome to apply for Beta access here.

There’s so much more more coming to Gigapixel, and we can’t wait to share with you everything else that we have been working on.

We’ve also fixed a number of important issues with this release of Gigapixel, as well as the previous patch releases. You can check out the full changelog below.


Dakota Wixom
Product Manager
Topaz Labs


  • Add patching overlay for the recovery preview
  • Added color-space and color management improvements
  • Add new Recovery Speed model
  • Improved the Low Resolution v2 model
  • Allow new models to be added to comparison view
  • Added collapsible and expandable file list
  • Added comparison view model elimination
  • Suggest Low Resolution version 2 over version 1 in Auto-mode
  • Removed low resolution v2 ‘Beta’ tag
  • Added model names in split and side views for capture
  • Increased checkbox mouse area size in open file list
  • Update dimension popup to display PPI and output size in inches/centimeters
  • Display dimension popup for file when hovering over its output size in the file list
  • Disable Recovery preview sizes that caused preview controls to go off screen
  • Disable “Close Image” on right click menu when comparing models
  • Updated capture button icon
  • Fixed panning in comparison view not working on Windows
  • Fixed zoom breakpoints not going below fit zoom
  • Fixed occasional rapid toggling between original and enhanced image with spacebar
  • Fixed export settings not scrolling
  • Fixed “renew license” text not being visible due to font color

Known Issues

Our team is taking a look at fixing these items below in a future patch release:

  • Some users are experiencing issues with monochrome and grayscale files
  • Some users are experiencing issues with “ghost faces” using face recovery on some images

I can confirm for myself that Low Resolution v2 really works much better. Recovery Beta also made it significantly faster for me, although as a representative of the “lower performance class” (3-year-old Acer Nitro 5 laptop with 16 GB RAM, Nvidia GeForce GTX 1660 Ti with only 6 GB), I would have liked even more acceleration. Anyway, it’s encouraging, Gigapixel 7.2.0 really brings more than just cosmetic improvements. And it still keeps the likable UI!

I tried using Gigapixel 7.2.0 to improve the shot of my occasional friend the Woodpecker (405x431 crop from a photo from a distance with an average compact Lumix camera). Gradual increase, at the beginning Low Resolution v2 and then repeatedly 2x Recovery Beta, up to 7260x5172. I am very satisfied (but I am not a professional).

I didn’t understand why the “Rendering Image…” window shows me “17s Remaining” and then “15s Remaining”. The “s” can’t be seconds if it takes more than 2 hours. Unintelligible information of no use to me. Importantly, instead of seven hours, it now took about 2.5, which is certainly quite significant for me as well.

I’m curious to hear from experienced photographers here about Gigapixel 7.2.0 for their work. Well, I am attaching a sample of my significantly improved (sharpened, smoothed, denoised, only with the help of Gigapixel, Detail 75, settings Quality) and blown-up Woodpecker. Such a beautiful bird, isn’t it?


Photoshop Automate TGAI. Stopped working in Grayscale mode!

There is a longstanding bug in Gigapixel and Photo AI. When face recovery is used and the image is upscaled and saved with the same name, reopening the file without closing Gigapixel will display a ghost head.

If the face recovery is activated when the image is opened again.

And here, I renamed a random photo of a dog, and dragged it on Gigapixel to open it :

Of course, closing the program and opening it again works, but it’s a bit annoying.
Most of the time, when I’m exporting my images, I just select Allow overwriting because I almost always make copies before doing any work on them. And I often re-import them to use another upscale model.


Hey Richard - are these images cropped by chance? This is something we’re tracking down in Photo AI as well.

1 Like

No crop.


Hey Richard, we were able to reproduce the issue, thank you for the steps!
Will look into it.


I won’t be able now but the steps are :

  1. Open Gigapixel AI (or Photo AI, I didn’t try again with it after the updates)
  2. Drag and drop a picture with a face (I used this one
  3. Select any model, and check “Face recovery”
  4. Export Image > Check : Allow Rewriting (so it keeps the same name)
  5. Remove the image from the list in Gigapixel with the trashcan icon
  6. Drag and drop the image that was upscale (which has the same filename)
  7. Activate the “face recovery”

And then you’ll see the ghost face.

That’s how I got the dog with the ghost face too, I just renamed the photo of the dog with the same filename and extension of the previously upscaled image.

It must be a cache problem somewhere.


I had this problem with Photo AI too, when the V3 came out, but I didn’t test this issue since


Those are some perfect steps, even the source image is included!
The ghost dog face is fun :))

It must be a cache problem somewhere,

It looks like that, hope to get it fixed soon!


Thanks Richard, we appreciate the help in chasing this one down. Please stay tuned.

Why is Gigipixel 7.2 taking so long to install? The installation is crawling along at 1 m per second. Been running for 4 hours and not even halfway installed.


When during installation the download of the 100 files gets interrupted the installation finishes anyway. How do I get the missing files after? Edit: ah, the application will notice when files missing and as soon as I start working a download progress panel appears. :slight_smile:

1 Like

“Also, please note that PNG files only support the sRGB colorspace in Gigapixel.”

Does this mean that colorspace is not being preserved for PNG? For how long has this been true?

If so, this is a bit of a shock and potentially wipes out a lot of work which must be now be redone from scratch at a later date.

Because when I choose an output colorspace from the list, or “preserve input color space”, I do so expecting it to be working?

And if it is not working, why is it that all the colorspaces are selectable for PNG in the export dialog rather than limiting it to sRGB if that is the only one that is going to work?

I was using the “preserve input color space” to process images that are Display P3, BT.2020 and BT.2100 PQ (HDR).

[This has really knocked me over. Am I going to have to review months of work to test whether colorspace was degraded? Am I going to have to redo multilayered piecemeal restorations from scratch? Is there any point in doing the rescaling work I had planned for this evening?]

This looks like it has been introduced in this release. The last 7.2 beta was still able to save png files in other colour spaces (TPAI still can), so not sure why this limitation has been introduced.

BUG: crop tool does not get applied when Recovery model is selected


Can you provide the system logs when this happened?

We can review it with our dev team.

How to Access System Logs

DeemChan was able to reproduce it :slight_smile:

Hi all,

Ran another test today. Although I ran comparison tests earlier to determine if the upgrades were satisfactory to me, my conclusion remains, I will stick with version 7.1.3. For now.

From left to right:
7.1.3. - Detail 100 Beta - Detail 40
7.2.0 - Detail 75, Face Recovery 20

Hi, there.
There’s been a bug with the Recovery model for 2-3 versions now. When it finishes rendering, it renders again as long as the first one if I have the misfortune to move in the image or if I touch the “Detail” slider. I remember that there was no problem in the first versions of the software with the introduction of this model.

It’s really frustrating
Especially as it took me at least 35 minutes of rendering time to scale this poster for restoration.