Photo Roadmap Update (April 2023)

[eric (Eric Yang)]

Product Updates

Since our update last March, we’ve significantly improved raw rendering and added a Preferences panel with new customization options.

Improved color rendering for many raw files

Topaz Photo AI will now display raw files more accurately for over 100 new cameras. For example, here is v1.2.10’s rendering compared to v1.2.8 for a previously unsupported raw file:

Depending on your camera, this change should fix color issues that you’ve previously seen when loading your raw files. Over the past few months we’ve undertaken a major project to improve raw rendering, and these results represent our progress so far. Please send us any raw files with remaining color issues and we’ll correct them in a future update.

New Preferences menu

We want Topaz Photo AI to fit seamlessly into your personal workflow, so we’ve added a new Preferences menu that offers two new options:

• Face recovery: you can now choose to apply recovery by default to all faces, only faces in the subject, or only low-quality faces. You can also now set default recovery strength.
• Close after saving: you can now choose to auto-close images after processing.

We’ve also consolidated all performance, Lens Correction, and Autopilot Configuration options into the Preferences menu so you can change them in a single place.

Also, updating Autopilot options in the Preferences will now immediately apply changes to all open images with Autopilot settings. (Prior to v1.2.7, updating Autopilot Configuration would ask you to restart to apply changes, which resets your workspace.)

We’ll continue adding workflow customization options in the future, so let us know which ones would be most helpful.

Other improvements

You can also enjoy a significant number of smaller improvements to Topaz Photo AI launched within the last month:

• New Open Recent option in File menu
• New ▲ / ▼ button on bottom-left that toggles thumbnail list view
• New bit depth options for PNG
• New shortcut for Close All Images (Ctrl/Cmd+Shift+W)
• Improved autopilot logic for subjects and multiple filters
• Improved authentication handling for various error states
• Improved monitor colorspace conversions
• Improved metadata handling in Lightroom Plug-In Extras workflow
• Improved Spacebar interaction with Original button
• Updated lens correction database
• Fixed minor memory leak when saving images
• Fixed downscaling being disabled due to Autopilot suggestions
• Fixed issue where folders with dots in them would not save on Windows
• Fixed Wacom tablet not releasing preview after panning

For a full list of release notes and weekly discussion, read the release threads for v1.2.10, v1.2.9, v1.2.8, and v1.2.7.

---

Feature Roadmap

We’re now actively working on:

• Adding a Remove Noise model for non-raw images with better slider responsiveness (beta).
• Adding a Strong Sharpen model that produces fewer artifacts (beta).
• Fixing occasional blurry patches in Enhance Resolution (beta).
• Adding a Refine Text filter that applies a mild deblur and removes artifacts from detected text.
• Adding a Straightening function into the crop tool.
• Improving preview panning performance by reusing already-processed tiles.
• Improving Autopilot decision-making for Remove Noise and Sharpen filters.

If you’d like to test features still in beta, please apply for our beta testing program. We hope you enjoy this month’s updates, and we can’t wait to get these new improvements to you.

Please let us know what you think by commenting below. You can also send sample images to our Dropbox or attach them to your reply if they’re under 65MB.

---

New Docs Page Release

Finally, we’re very happy to announce the launch of a new Docs page for Topaz Photo AI.

These pages are just the beginning for making Topaz Photo AI an easy and engaging experience users. We welcome any feedback and suggestions on how to improve this new section of our webpage, and we will be keeping our users and beta testers involved in future updates.

Thanks for using Topaz Photo AI!

[Kadet (J.W.)]

@eric

Adding a Refine Text filter that applies a mild deblur and removes artifacts from detected text.

I’m waiting for this in TVAI

[christian_p_schaefer (X)]

@eric

Fixing occasional blurry patches in Enhance Resolution (beta).

Are you also actively working on the „Titles Issue“ we spoke about?
(inconsistent denoise effect on certain grainy images, some out-of-focus tiles get more sharpened than others)

[yunowe69]

When will the command-line interface support upscaling?

[Laco]

If Photo AI is to be the new all-in-one image enhancement tool, shouldn’t it include a full set of colour correction tools and a heal tool?

My understanding is that Topaz have stopped developing Studio 2 because the company wants to focus more on image enhancement and correction, rather than image creativity. I too, mainly use image software for enhancement and correction purposes; but the most common things I correct, apart from sharpness and noise, are colour (saturation, temperature, brightness, etc.) and spots and blemishes. For these I have to use another program in which there is a range of colour correction tools and a good heal tool (for the blemishes).

At first mention, a heal tool may seem quite a way off the roadmap; but if you consider that when Photo AI enhances a low resolution image, the result is often successful except for the introduction of one or two anomalies, this is where a heal tool would come in handy, within the program itself.

For example, I upscaled an old photo recently and Photo AI did a very good job of enhancing everything to a state of greater clarity, but in so doing it did unfortunate things to one person’s front teeth. Just because of this one anomaly I had to import the result into another program and edit the teeth with a heal tool. It would have been great if I could have done that within Photo AI.

Denoise, sharpen, upscale, crop, colour-correct, heal—a more comprehensive image editing package, still with the focus on enhancement rather than creativity.

[Parker]

Disappointed.

When you bring in a file to be a JPG or TIFF, PAI recommends… “to maximize raw image quality process the raw file with the direct plugin.”

However, at this time, a RAW workflow is useless. I brought up several issues with RAW files weeks ago and am sad to see it’s still not on the Roadmap to be fixed.

Fixing current issues should be a priority over “new” features.

RAW workflow issues: masks are not being applied properly (and/or missing masks), crop via Lightroom is not being carried over, and the noise/sharpen models for RAW are horrible.

[dwalme]

What would be great to see is Photo AI saving settings for each file that is processed with the tool. So if files need to be reprocessed in the future you start where you left off.

I processed an entire folder of images and while editing the DNGs in Capture One I realized that I missed the fact that there were artifacts in some images and I had to go back reprocess them.

It would be nice to load that folder of files and have all the settings that were originally used loaded at the same time. By being able to see what corrections were used it would help narrow down the source of the artifacts. Then I could go through and quickly change what need to be changed without having to start from scratch.

[bobsun (Robert)]

Any thought about somehow saving image process parameters that are lost on closing application? Ideally they would be added to the exif file. A simple process log file for the app would suffice.

[Fotomaker]

All of the on schedule Roadmap features sound useful (some I’d use more than others b/c my workflow is in Photoshop vs using the Standalone PAI version).

As the currently-detailed upcoming features get checked off (ie, ready for prime time & release) other additional features I’d wish were high priorities to be developed for us Ps Plugin users are:

• Add Support for the Gigapixel (i.e., Enlarging) part of the PAI triad in the Plugin. If that means altering PAI to run from the Ps Automate menu vs Filter menu that’s okey dokey with me. I use enlarging a lot and it’s very inconvenient now to have to break up that part of my image processing workflow and use PAI enlarge outside of Ps. I instead still rely on an older version of GAI inside Ps…

• Add the ability to Invert Masks in PAI (both Standalone & Plugin versions). This is critical for times when only a small portion of the image isn’t meant to be the masked subject - we should be able to paint that small area and invert so the bulk of the image we want masked can be done so much much faster than now.

• Related to my 2nd bullet - but lower on a priority level for me - add a classic Topaz filters feature from the ‘good old days’… the “Apply” Button inside the Masking UI. So we can mask/paint one area for PAI treatment, make that treatment (ie, settings adjustments), then go back into the Masking UI, hit the Apply button, then paint a fresh mask for a new treatment onto the image almost as if it’s a 1st mask - but is done in a single processing session. This would definitely help us one-at-a-time processors, but may or may not be useful to batchers too.

Thanks!

[michael.gale]

Hi Topaz,
Thank you for the autopilot menu! It’s soo much faster now.
An even bigger thank you for asking for feedback on workflow customisation options. This is something I am very passionate about so I would also love to give some ideas on how this could be updated really expedite workflow (if it’s not already on your dev list).

Quick wins

1. Add “Save all images” to menu (with autopilot settings. That way it can be called by apps like Keyboard Maestro)
2. Remove the confirmation button when you want to apply settings to all.
3. Remove the check of the settings when you click “save all images” just set this in autopilot settings and go.
4. Add the following to the autopilot preferences:

• Auto-upscaling the output size to have additional options of “shorter side” and “longer side”
• DPI output
• Filename, file type, suffix/prefix and output folders.

Bigger Items

1. Autoprocess - Have an option to automatically process and output the files it receives. This could be controlled maybe by an off switch in preferences.
2. Watch folders - when a file is dropped into a folder it is automatically processed.
3. Autopilot Profiles - Update the AutoPilot to be able to create more than one “profile”. This would then be able to set all the things in autopilot plus the items mentioned in quick wins. Additionally, you could assign a watch folder to each profile. This way you could have separate folders setup with different profiles attached which work for different artwork.
4. Command Line - Further automation by integration into command line use, so it can be utilised by apple/os/keyboard maestro and automatically applied in the background and merged into a larger automated process.
5. Larger upscale so we don’t have to upscale twice for large images.
6. Photoshop file (although this would then need to processes layers and masks etc so I’m guessing this could be huge)

Thank you for your time and thanks for making such an AMAZING app!

[kevin.shank-0439]

I noticed tpai.exe cli version echoes the auto settings and generated/manual parameters so there is already a format that will work for such a setting file, just needs to save/read it! It would need the option to read only the auto settings to regenerate parameters or keep manual parameters though.

[Kenny (Kenny McLean)]

What you’ve done so far is amazing.
I know that text is on your roadmap

• Adding a Refine Text filter that applies a mild deblur and removes artefacts from detected text.

At the moment Photo AI is rather hit or miss. When it gets it right it does a decent job. When it gets it wrong, though, its diabolical.
I’m using Photo AI at work to resize web graphics for a Shopify site.
The logos, images etc that I grab from the manufacturers sites are usually low res. Photo AI often allows me to upsize 2-3 times to get them big enough. It doesn’t seem to recognise text, particularly when it is angled, or curved. One example, which unfortunately I didn’t save, had me bursting out laughing it was that bad. It was like it was written in Klingon or Elvish or maybe Photo AI ish.

[iMacmatician]

I’ve found that the face detection, when set to Auto, will often activate Face Recovery on faces that are of medium quality and are not so small. In these cases I find Face Recovery to generally be worse than no Face Recovery, especially with keeping detail.