Product Direction in 2024

Hi, and thanks so much for trusting our products with your work over the past few years.

In 2024, our photo and video products will experience the most significant evolution since our first AI product six years ago. I wanted to share what you can expect out of our products in the next few years.

We believe that generative AI should improve your work rather than replace it, so our product development philosophy focuses on three principles:

  1. Enhancement vs replacement: for improving your work without altering its core essence.
  2. Natural and photorealistic results, vs fantastical: for your photos and videos where preserving reality matters.
  3. On-device processing, vs cloud: for a faster workflow, more editing options, and private processing.

We think that these principles are essential for professional use cases, but currently undervalued in AI research. As such, we’re accelerating this direction with our three core products:

Photo AI
Photo AI is designed for photographers, and solves many photo quality problems like noise reduction and sharpening better than anything else. In 2024, we will use generative AI to greatly improve the quality and experience of other photo editing tasks.

Video AI
Video AI delivers you sharp, steady, and smooth footage without motion artifacts. In 2024, you will see improved AI model quality, a better app experience, and more integrations for professional editing workflows.

Built for everyone, Gigapixel 7 brings improved upscaling model quality and better usability. Later this year, we’ll release fundamental improvements to our upscaling models that unlock another level of quality, better than anything else you’ve seen yet. Afterwards, we will bring Gigapixel to more platforms than just desktop.

We’re incredibly excited about this direction, and you’ll soon see some of these improvements in the next few months.

We’d also love your feedback as we’re filling our roadmap for the rest of the year. What would you love to be able to do with your images and videos that you can’t currently? Is there anything you wish our current products would do better? We will read and consider every comment.

Thank you again for trusting our products, and we will do everything possible to continue earning that trust!

Eric Yang
CEO, Topaz Labs


I use Photo AI pretty much daily. I regularly enter competitions. Something that has come up is that competitions are moving to ban generative AI. It would be good for me if it were clearly indicated what uses generative AI and what doesn’t.


Nice writeup Eric, I’m very excited for 2024 and beyond!

This is a pie in the sky idea/wish that is probably a few years off, but for me the ultimate use of AI for video would be to use generative AI to convert 4:3 content to 16:9/wide, filling in those black pillarbox areas with content that passes for ‘believable’.

I can expand a single frame of a video with Photoshop’s generative expand which takes several seconds and isn’t even using local processing but the Adobe cloud servers. To do an entire video would require multiple times the processing power of a RTX 4090 to even be feasible. So, maybe by the time we get to the RTX 7xxx series cards it will be a reality.

But, until then, imagine a couple of old TV series, brought back to life in widescreen formats that never even existed:

Magnum PI:


Of course there would be additional challenges like how to keep consistency between frames and scenes and even episodes for all the generated content. Like I said this is a very pie in the sky idea that I don’t expect to see for several years.

In the meantime, I’m very happy with seeing the steady improvement of the models in Photo AI and Video AI.

Hi-Fi v2 in Photo AI and Proteus 4 in Video AI really gave me that ‘wow’ factor for the Topaz products again - and I hope things can be improved upon even further!



It will be great if it can do Sharpening without generating Halo defects.


Turn on Hifi V2 in 1:1 resolution, best in Photoshop on a layer to control the strength via opacity

One of the things that has made Photo AI and Video AI so invaluable to me is their ability to improve images when I couldn’t fully express what was wrong with them in the first place. As such, I’m not entirely sure how to describe what I need. Like david.123’s reply, mine is somewhat pie in the sky.

I have been using Video AI to bring old video footage shot on 8mm, Hi8, and MiniDV up to 1080p and occasionally 4K.

Unlike david.123, I don’t really want my 4:3 pillarboxes to be “filled in” with generated content. However, I would very much like the quality of what was recorded to be improved upon in a way that brings it closer to looking like it might have, had it been captured with a modern camera.

I know when viewing these old recordings, the human brain does a lot of work to compensate for missing information. While traditional methods of video enhancement could only utilize data that exists in the original image, I feel that we are approaching a point where AI could do more compensatory processing and not only allow for faithful increases in resolution, but actual improvements to image quality and clarity.

I hope that makes some kind of sense. I’m happy to share some of my footage offline if that would be deemed helpful.

In the meantime, I’m excited about what you guys are bringing in 2024!


Hi Eric!

I’m Richy, and it’s finally a pleasure to meet you. As a longtime fan of Topaz’s software, I’ve always been fascinated by the brilliant minds behind it. It’s great to finally associate a face with the name.

I absolutely love Photo AI, and it has become an essential tool for me. I’ve been using the limited free trial for over a year now (not sure how I feel about admitting that, haha!). I’m planning to purchase the entire suite soon and am excited to learn more about Topaz’s plans for the coming year.

Your practical approach to AI really sets Topaz apart from the competition and will only solidify its unique position in the market.

This old Coca-Cola adage comes to mind: ‘Often imitated, never replicated.’

Cheers to an incredible 2024 for Topaz!

Take care,

1 Like

Hello Mr. Yang,
The software is amazing when it is used correctly and within reason. My request to you beyond what you already suggested, is to add several more templates for specific common conversions as starting points. I think it is reasonable to ask for templates optimized to convert files from 4:3 VHS and files that have been digitized from old 8mm film by popular devices such as the Wolverine devices. Create some good suggested templates for these formats which are pretty common jobs for us doing digitization and restoration work.


Exactly! Real world scenarios that need to be addressed with some good templates. Maybe even add a mode optimized for us archivists who are digitizing 4:3 video VHS/Beta, and 8mm film reels.


Upscale and export gif in photo ai😊


Other optimizations that need to be addressed for 4:3 VHS/Beta and 8mm film projects is better temporal noise reduction (like NeatVideo plug ins), maybe a “de-skew” parameter to fix video that has “flagging” at the top of the video, maybe an automatic setting to sense where the frame of an 8mm file is, and crop it appropriately. Maybe a parameter to fill in the “sprocket hole” on the left side of an 8mm film frame, maybe an auto-leveler of the brightness (8mm film tends to flicker from bright to dark due to inconstencies of the exposure of the film), maybe an intelligent auto-leveler of the white balance (8mm film tends to change at random the white balance due to film quality inconsistencies), maybe a simple speed adjustment that easily allows us to adjust the frame rate (digitized 8mm film files are often not the correct frame rate.)


Hi, I’m a fan of Topaz’ efforts so far, though I haven’t used them with success yet, I’m just happy to support good ppl doing good work in this direction. I know the day is coming when the Ai will be good enough to make something of all the footage I’ve digitized. I’m specifically waiting for one thing that’s going to make a huge leap forward. When we can feed the model multiple higher quality still photos of a subject (bc photos were so clear compared to 8mm or vhs frames), and then tell the program, “These stills are the face of that person in the video.” and have it basically deepfake them in place to match the movement in the scene. Further, here are the other frames of this persons head from in this film as the camera moved around them at various times and gave us closeups. Use them to build a higher resolution model of this persons head. To go even further, here are some high resolution pictures of a black T Shirt, apply that to this bassist, here is a black fender Stratocaster, vintage WEM cabinets and here are high resolution shots of the ampitheatre in Pompeii, put it all together, and we can finally build a clear version of a lost original film. This is the dream. I will be waiting patiently with my archives.


Hello, Mr. Yang.
I’m new to Topaz products but I continue to be surprised and impressed with the quality, functionality and simplicity of Topaz software. Your request for a wishlist is timely as there are a few additions or enhancements that I would love to see.
I do underwater videography, primarily macro. Lighting is difficult with backscatter (reflection on suspended particles,etc.), motion is a constant battle, and opportunities to take the shot are limited. Wet environs have their own challenges with focus and lens correction. All of these lend themselves to AI modeling and specific presets (IMHO). This sounds specific to the underwater world but action and nature focused footage could benefit as well.

On a second, and perhaps simpler note, I need an app on my iOS device that would give me Video AI queue status updates. Did the export run to completion? Did it error out? Estimated time remaining? Without the luxury of a server farm or very high-end hardware this can be a frustration point and a productivity issue on longer segments. Even a text would be helpful.
I’m looking forward to the continued success of Topaz’s amazing toolset.

1 Like

If you read the opening post it tells you Photo AI will use it this year.

Generative content will no doubt fill the planet with more imagery… however, tools will still be needed to improve traditional photography.

Towards that, a tighter integration with photoshop will be a real boon… for example, if GigaPixel could be a layer- that can be controlled at anytime, one could go back and improve or reduce sharpening as the edit progresses. Non destructive editing of effects, and also selective editing improvements would be superb.

Towards AI - if we coulfor example circle the area and give it certain commands like, balance the Color or blend luminosity to be within a certain range or such means of getting help to do actions. Select hair in this area and improve it fill in details.

the ability to brush in a certain artistic effect… for example a can we have an AI VanGogh brush that you could dial in the intensity of the effect and paint with it over the image

Thank you for the great products and the roadmap announcements. As a long-time user of Topaz products, I am also pleased to be able to use innovative algorithms for image enhancement offline. However, I would also like to see the introduction of a creativity slider, especially in the upscale software, which generates AI-generated details and structures on request. I’m now increasingly using the highly acclaimed Magnific AI, which takes this approach and delivers great results when the source material doesn’t have enough detail. Such technology would certainly add tremendous value to Topaz products and encourage many enthusiastic MagnificAI customers to switch to Topaz. I look forward to the future and hope for great innovations.


I see this technology as the only way to further improve enlargement.

I can’t imagine it any other way.

Denoising is another topic that has become very quiet, but I think that’s because it’s being worked on.

I would like to have natural, high-resolution images, with lots of details, precise colors even with high levels of noise, which can be good edited afterwards (without artifacts, without loose of details), without cuts in the dynamic range either in the shadows or in the highlights.


Denoising is a complex issue because, unlike sharpening, it depends on the way an image is taken. There are so many different combinations of sensors, pixel sizes, atmospheric conditions, lighting conditions and, of course, the operators skills.

Maybe the simplist solution is with Generative AI?

Thank you Eric, for your encouraging words.

I use Gigapixel regularly via Lightroom, but I really miss the ability to use it in the export window of Lightroom. At the moment, you need to export it to gigapixel, open the images, and then select the options. I would be happy to have the settings in a Lightroom plugin, or if that’s too complicated, set on standard auto and press one button in the Lightroom export window to get the enlarged file without opening Gigapixel.

Regarding file naming, I would be happy to have the files the same as the Lightroom template. I don’t need to add a suffix to them.

The whole process could or should be seamless, like an API or a Photoshop action that runs in the background or even executes gigapixel in a batch mode via Lightroom with the parameters that I set in the Lightroom export window (gigapixel plugin).

At stage 2, add more of the settings to the Lightroom plugin, but you can start with the minimum, jpg 80%, enlarge x 2, standard, default settings, and no file renaming.

Here is how I see it.

Thank you

1 Like

I would like to see an option similar to where prompting can be taken into consideration in the upscale of an image, fundamentally changing or altering the style or substance of an image if requested.

1 Like