In the last several days Topaz has released updates to its new versions of both Photo AI and Gigapixel AI.
The details as presented by Topaz are pasted below.
But first, some of my own examples of old family photo restoration using the latest releases.
PHOTO AI
A section of a scan of a small snapshot with Upscaling, Super focus and Face recovery (even the dog!):
A screenshot of an online yearbook photo with Upscaling, Denoising and Face recovery:
A tiny portion of an 8×10 contact sheet of several box camera negatives, with 4X Upscaling and Face recovery:
Face recovery and 4X upscaling of a face from a single box camera negative contact print:
A 2X enlargement of a portion of a 1990s scenic from a full-page scan of a photo album page:
Photo AI v4.0.0
Released May 8th, 2025
Dust & Scratch Repair: It’s finally here. You can now resurrect old images by healing major scratches and dust particles for a clean image using the world’s best AI dust & scratch model only in Photo AI. It runs both locally and in the cloud, comes with a built-in healing brush for finishing touches – and it’s going to get even better over time!
Revamped Autopilot Experience: Autopilot is more intuitive to use than ever before with this update, and you’re going to love the refreshed design. Booting up Photo AI for the first time will show the option to opt-in or out of Autopilot, and you can change this toggle at any time using a handy switch on the right hand panel. Suggestions also appear at the top of the filter stack, and are much easier to work with.
Neck and Hair Controls for Face Recovery: More control over faces. Face Recovery improves details on faces, but also on hair, necks and shoulders – but we’ve never exposed these controls until now. Get the best Face Recovery results possible with more control over the model with this update. We’ve also split the controls for Face Recovery strength and selection to make it easier to work with in this release.
Dynamic Face Recovery Strength: Get better faces every time. Autopilot will now use the best Face Recovery strength settings based on the size of the faces in your image. This creates much more natural results, preventing over-processing particularly for larger faces. Autopilot will also now ignore faces in the background which shouldn’t be processed, giving you better, more consistent results every time.
Autopilot Intelligent Selection: Autopilot now senses whether your image may be a landscape, a portrait, or something else – and automatically updates your default selection to get the best results possible. This change is going to dramatically improve your Autopilot results for large batches of images by focusing on the most important subjects in your image.
More Improvements: There’s even more under the hood. Denoise and Sharpen controls have been improved, and filters are now reverse chronologically ordered with the most recent filter appearing on top instead of the bottom of the stack. We’ve also improved the login flow, adding SSO options, and we’ve fixed many bugs to improve your experience with this release.
Photo AI v4.0.1
Released May 30th, 2025
Changelog
•Add personalization on/off toggle
•Added cloud processing enable/disable toggle
•Add animation for healing brush processing
•Removed beta tag for face recovery v2
•Remove selection from super focus before processing occurs
•Fix incorrect text color in dust & scratch modal
•Fix sharpen model grid select alignment
•Fix denoise value set below 1 breaking processing
•Fix invalid login error for existing account
•Fix recover faces manual strength not applying to images
•Fix clicking off license dialog cancelling login
•Fix healing brush and remove processing conflicting when used on the same image
•Fix original image flickering before processed image appears
•Fix preview status for dust & scratch processing
•Fix export with wrong file name failing to export
•Fix Apple Photos plugin export hanging with some file types
•Fix error code when using LrC > File > Plug-in Extras on Windows
•Fix monochrome non-raw output having 1 channel and failing to import to Lightroom Classic
Gigapixel v8.4.0
Released May 22, 2025
We’ve enhanced how Gigapixel processes your images with the introduction of Pre-downscaling, now seamlessly integrated into the Recover v2 model.
When working with images that have been previously upscaled or contain limited detail, results can sometimes appear inconsistent. This often stems from low information density—where image size has increased without a corresponding boost in visual detail. Pre-downscaling addresses this by intelligently optimizing image input before enhancement.
How it works:
Pre-downscaling: Smarter Input for Sharper Results: Pre-downscaling reduces resolution to re-concentrate information density.
The enhanced input is then upscaled using AI to restore natural sharpness and detail.
Choose between four different intensities to preview the intensity of the downscale.
Personalized Adjustments (Auto Mode Learning): We’ve enhanced Auto mode with Personalized Adjustments, which learns from your export history to fine-tune Auto defaults. Applies to Detail and Texture settings (excluding creativity levels in Redefine modes).
Enabled by default.
Expanded Language Support: Gigapixel just became even more international. We can now support the following languages: German, French, Spanish, Portuguese, Japanese
Redefine Evolved, Two Powerful Paths: Realistic & Creative Modes: The Redefine model now is split into two tailored enhancement modes to effectively move through your workflow. Realistic mode is best for enhancing and refining existing details in an image, while Creative mode is better suited for reimagining or altering those details to achieve a different look.
Cloud Compare Tool: Easily view, compare, and share results from generative models with the Cloud Compare tool.
One-Click Support: Get Help Button: Borrowing from Photo AI, we’ve now added the Get Help button directly inside Gigapixel. Instantly sends diagnostic data to our team. No email or log hunting required. Fast, frictionless, and user-friendly.
Changelog:
•Added new pre-downscaling option to Recovery V2
•Use this when enhancing naively upscaled images to get better results
•Added app translations for French, German, Japanese, Portuguese, and Spanish
•Redefine is now split into Realistic and Creative modes
•Added personalized adjustments for autopilot
•Status will now show if a generative model is currently preparing
•Added cloud compare button to finished images in cloud queue
•Added new Get Help button to report issues
•Updated Exiftool to version 13.27
•Improved model verification check logic
•Fixed various misc. crashes
•Fixed logging out not clearing token correctly
•Fixed some issues with model downloads in installers
•Fixed face recovery sometimes turning off in batch exports
•Fixed unselected images being visually added to the export queue
•Fixed various focus issues
•Minor UI changes
•[PRO] CLI now supports generative models
•[PRO] CLI should now suppress more internal logs
•[PRO] CLI should now not get stuck after finishing on some machines
Gigapixel v8.4.1
Released May 28, 2025
This release brings a round of improvements and stability focused fixes following the recent 8.4.0 update. It focuses on tightening up face detection and recovery workflows, improving input handling, and refining some of the internal processes.
Changelog:
•Changed face recovery auto selection mode to select high confidence faces
•Fixed models failing to download or verify
•Fixed face recovery settings causing auto face selection to fail
•Fixed RGB images tagged with Grayscale ICC profiles not loading
•Fixed close image after save functionality when not in English