Review Of Topaz Photo AI: A Single App With A Variety Of ML-Powered Tools
Applications like Adobe Photoshop and Lightroom, DxO Optics Pro, Capture One, and others give you access to a wide range of controls with which to customize your artworks to your heart’s content. But what if you have a large number of photos to edit and just need good results quickly? Topaz Photo AI, Topaz Labs’ newest product, fills that need.
Instead of focusing on complexity like most photo editing apps do, Photo AI aims to handle the tedious work for you. This is made possible by a clever Autopilot feature that examines your images, determines their subjects and any image quality issues, and then uses this information to apply intelligent corrections with the least amount of user input. Topaz’s well-liked DeNoise AI, Gigapixel AI, and Sharpen AI applications provided the underlying technologies.
- Works with both pre-rendered JPEGs and raw images.
- DNG Raws for further processing or finished JPEG, TIFF, or PNG files are produced.
- Can be used as a standalone application or as a plugin for applications such as Photoshop.
- locates your subjects and instantly detects image quality issues
- eliminates jitter, blur, and lens distortion
- AI algorithms are used to upsample low-resolution images to higher resolutions.
- finds faces in your photos and enhances their visual appeal.
- Using the autopilot tool only requires modifying masks and adjusting strength sliders.
- No membership is required.
Topaz Photo AI costs $199 and is currently accessible for Windows and macOS. With a more efficient workflow on top of it, that represents a $60 saving over Topaz’s standalone AI-powered apps. Let’s examine the features that Topaz Photo AI offers!
A clear and uncomplicated interface with a few oddities
Topaz Photo AI has an incredibly straightforward and tidy user interface, which is to be expected given its fuss-free approach to editing and complete eschewing of features like photo management. It has two modes of operation: standalone and plugin for other programs like Adobe Photoshop. When the app is launched independently, you can add images by dragging and dropping them onto the app or by browsing your drives and selecting them to add either individually or in groups.
Once an image has been loaded, Photo AI will check it for objects, people, and poor image quality before automatically generating the adjustments it thinks your picture needs.
A filmstrip of all your open images is displayed below a preview of the current image in the standalone app. However, oddly, the filmstrip will only display JPEG image thumbnails. Even after opening and previewing Raw files in full size, thumbnails are not produced; this is something I’d like to see changed in a later version.
Approximately 1,200 cameras, camera backs, drones, phones, tablets, and other allegedly compatible modern devices are included in the raw support, which is quite extensive. (The most notable omission is for the ‘High Efficiency’ Raws’ of the Nikon Z9, for which support is anticipated but not yet available.)
A panel with an image navigator and only a few controls to adjust the automatic corrections can be found at screen right. You can access the mask, brush, and zoom controls from the menu bar at the top. You can also change preferences, such as whether the program should primarily use your CPU or any available GPUs.
Strong automatic subject detection, but it can be time-consuming to refine masks
When Topaz Photo AI has finished looking for subjects and faces in your photo, you can check its findings by moving your mouse over the subject and face fields in the Autopilot. Each detected face is marked with a yellow box, and the subject is highlighted in red on both the image navigator and the actual image.
I was pleasantly surprised by how accurate the face detection was, identifying faces even when they took up very little space in the image and were partially or completely hidden. Despite the lack of a manual selection option, you can deselect any faces that the algorithms identified that you don’t care about.
When the subject of the photo is obvious, as in portraits, the subject detection does a pretty good job, but with things like landscapes or street scenes where the subject doesn’t stand out from its background as clearly, its selections can be a little more hit-and-miss. In these circumstances, you can manually change the mask or switch the algorithms from portrait to landscape mode.
But rather than directly reshaping your mask with a brush, you choose from four different chunk sizes to add or subtract from “chunks” that the algorithm has identified in the image. This works fine for minor adjustments, but if you need to significantly alter the subject area, it can become laborious. I found myself wishing that there was also a fully manual refine brush available.
Although it reduces work, autopilot is missing some crucial features
After identifying subjects and faces, the Autopilot algorithms begin to analyze the image for flaws like noise, soft focus, motion blur, and low resolution before suggesting adjustments that are appropriate for the image. Additionally, automatic lens distortion correction is applied by default but can be turned off if you’d rather not. Overall, Autopilot is very user-friendly and effective.
However, there is no way to fix exposure and white balance problems. In the aforementioned example, I used the same underexposed, cool-tinted image as a starting point in both Photoshop and PhotoLab, but I was able to fix it in a matter of clicks. Contrarily, Photo AI won’t automatically correct the exposure or white balance for images that have significant underexposure or overexposure, an extremely wide dynamic range, or blown color casts. Additionally, there is no way for the user to manually fix these issues.
Since this is the program’s biggest flaw, I would really like to see it fixed in later iterations. It means that, for the time being, Topaz Photo AI must still be used in conjunction with Photoshop or a program of a similar nature. But if your shot has good white balance and an accurate exposure, Topaz AI does a great job with almost no effort.
With almost no effort, excellent and quick noise reduction is achieved
The noise reduction algorithms developed by Topaz Labs are probably the company’s most well-known product. These algorithms are also offered separately (and with slightly more user control) in the standalone DeNoise AI app.
The DeepPRIME XD algorithm from DxO PhotoLab produces results that I find to be slightly superior, but there’s no denying that Topaz’s most recent algorithms are very good indeed and much faster than the extremely processor-intensive DxO algorithms. Even from extremely noisy images, the Raw Remove Noise AI tool in Photo AI can produce surprisingly clean and usable results while still doing a good job of preserving the fine detail.
The defaults, in my opinion, can be a little too strong on noisier images, giving them a slightly “plasticky,” over-smoothed appearance and introducing some noticeable artifacts, especially in repeating patterns. However, you can tone things down with the strength and detail sliders or just change the preset from “strong” to “normal” to get more realistic outcomes.
Although Topaz AI will occasionally lose some of the finest details to the algorithms, if a shot doesn’t actually require noise reduction, it’s worth turning this feature off completely.
Unexpectedly effective motion and lens deblurring
With the right subjects, I actually found Topaz Photo AI’s deblurring to be even more impressive than its noise reduction, though there are still a few limitations.
There are three different sharpening options: Lens Blur, Motion Blur, and a general Standard mode. As with noise reduction, there are only two sliders available to control the effect’s intensity. Additionally, you can decide whether sharpening is applied to the entire image or just to objects that were manually or automatically detected.
Although, as is typical of AI algorithms, they struggle most to convincingly regenerate fine text, which is something we humans are much better programmed to recognize, the algorithms can make a surprisingly blurry shot potentially usable with motion blurring. Additionally, they exhibit some confusion in situations involving complex motion, such as when a car’s wheels are moving in an opposite direction to the car itself for the majority of their circumference.
For instance, when looking closely at my race car image below, the sponsor decals appear a little wonky and the wheels appear to be out of round, but otherwise, the cars look much crisper than in the extremely blurry original image. The improvements are even better for shots with subtle motion blur, like the cycle race below, where even small text appears noticeably crisper while still being completely readable.
Topaz Labs aimed to create Photo AI, a program that can fix common image flaws like low resolution, noise, and blurring from defocusing or subject motion in a very user-friendly yet reliable way with little to no user involvement. And in many ways, it really does a great job of the job.
Its deblurring and resolution-enhancing tools, both of which accomplish quite difficult tasks with only a few mouse clicks, particularly impressed me. Although I found the face enhancement technology to be more sporadic in practice, when it works, it has a seemingly magical quality. Not to mention the noise reduction. Although I don’t think it’s quite as powerful as that from competitor DxO Labs, it still performs a great deal better than many competitors. (And it completely outperforms in-camera noise reduction.)