Right on right on sing it. AI “art” is starting to stomp on the soul of humanity!
The reason I haven’t gone for any paid updates is that Gigapixel AI doesn’t usually feel the need to sharpen or upscale my very abstract images. What I do is unique, sort of a 2020s evolution of modern art painting, only instead of canvas I’m using a computer to mold and “paint” with fragments of those images.
Point being, for it to be useful to me it would have to understand that the pixelated edges in a drive-by iPhone photo of, say, zoomed-in rust on a trash can want to be sharpened.
Shorter: to be honest, I haven’t found it to “upscale any image like magic with AI.” But I’ll be an eager customer again if and when it does.
This. On the continuum between algorithms operating on nearest neighbor pixels and high-level generation of entirely new content, I’m currently more interested in the former.
Many of the images I’d like to scale are scientific images that are quite niche. It’s highly unlikely a ML AI could have been trained on similar content, because there isn’t any. Given that, I’m less interested in hallucinations, however creatively pleasing they are, than low-level denoising, sharpening, and upscaling capabilities.
Editing my private photos as a hobby, I like Photo AI for real life photos. The results are often good. A big challenge still is, to enhance scanned photos which have a silk screen (either for fingerprint or copy protection). While Photo AI identifies the faces and really makes them look natural, the rest of the photo still has lines. If Photo AI would offer a specific filter or ai recognition and enhancement, it would be a huge step to digitize old photos.
Since I bought a license for Photo AI, there were constant updates. Sorry, but I didn’t saw the improvements, while I really trust, they are there. I still used Sharphen AI and Denoise AI more than Photo AI for both specific tasks.
I don’t get it, that the extension of the one year updates should cost 60% of the original price. Though I like the filter, it’s a bit too much after one year to me. If one uses the program professional, it’s not much money, though. Maybe, Topaz can convince me with a killer feature (hint: see beginning of my note).
Eric, I can’t say enough about the great technology your company and staff have done of advancing the technology over the years, its my go to program for Noise, Sharpeneing, and Upscaling. I also use Topaz Studio 2, PLEASE DONT FORGET THIS EXCELLENT PROGRAM, an update would be most appreciated and I think many people dont realize it even exists, so a promotion of Studio should be included and shown what it is capable of. All in all your software engineers are the best keep up the great work
Something that would be useful in Video AI is a way to reduce ringing effects. I work a lot on video footage from analogue recordings. You can spot the ringing really easily on say on a flag pole, where there are one or more ghost versions of the pole to the right or left, steadily reducing in intensity. You’d hope that AI would be the answer to this, learning what we humans can recognise easily, but I don’t think I’ve seen any of the models able to reduce this effect.
I have been a long time user of Topaz products, particularly Denoise AI and Sharpen AI. I have continued to try Photo AI since its release. Though Photo AI gives fairly comparable results to the combination of Denoise AI and Sharpen AI, the speed of processing is dramatically different. I am using a MacBook Pro (2019), with 2.4 GHz 8-Core Intel Core i9 processor, AMD Radeon Pro 5500M 8 GB graphics and 32 GB 2666MHz DDR4 Ram.
Here is a list of processing times for 2 images, first a Nikon 24 megapixel NEF raw file and the second an OM Systems 20 Megapixel ORF raw file.
Nikon 24 megapixel NEF file
Photo AI (RAW from File Menu) - 3:50
Photo AI (from edit in Menu) - 3:40
Denoise AI - 1:00
Sharpen AI - 0:44
DxO PureRaw 3 - 0.55
OM Systems 20 megapixel ORF file
Photo AI (RAW from File Menu) - 3:25
Photo AI (From edit in Menu) - 3:50
Denoise AI - 0:38
Sharpen AI - 0:34
DxO PureRaw 3 - 0:24
It takes less than half the time to process a file with a combination of Denoise AI and Sharpen AI (Or a combination of DxO PureRaw 3 and Sharpen AI), than it does to process with Photo AI. I have tried this test with multiple versions of Photo AI, and this test was done with version 2.3.1. I used the default Standard recommended settings for both Denoise Ai and Sharpen AI, and the default Autopilot settings for Photo AI.
Also, the results with the combination of Denoise AI and Sharpen AI (or DxO PureRaw 3 and Sharpen AI) are superior to Photo AI in most cases. It seems as though you are not using the same algorithms for Photo AI that were used in Denoise AI and Sharpen AI.
Though I have enjoyed using Topaz Products, the time penalties to use Photo AI, and the fact the the results are usually inferior to either of the combinations that I have discussed above is disappointing. With the apparent abandonment of upgrades for Sharpen AI and Denoise AI, I will be looking at alternatives to Topaz Photo AI. If you can somehow improve the speed of Photo AI, and match the results to the standalone Denoise and Sharpen AI programs, then I will reconsider. Nearly 4 minutes to process a 24 megapixel RAW file is unacceptable.
On a different note, the progress bar in Photo AI is misleading. Make the bar longer and only need to go through the length of the bar once while the file is processing.
I use Topaz daily as a web developer to improve low quality images that I usually pull off of clients’ old sites.
I’d love to be able to remove background and create transparent background images. I use remove.bg many times a day and if topaz did that I’d use it almost exclusively for my needs….
The only other thing to make it perfect would be the ability to export to webp / and more exporting options like the export size in pixels and quality.
I am a professional printer serving the New York market serving many We often print Chromogenic prints in the 50 inch range and pigment prints in the 60 inch range. The generative element will be useful to solve noise and resolution problems. My job is going to get easier. Less local masking, more time to spend with clients while being able to think more about color and contrast and aesthetics.
Cut down the price for annual upgrades - it’s ridiculous, especially for people who only use the products on an occasional basis. I won’t be upgrading unless the price comes down to something intelligent - like $50 or so. Don’t treat your loyal clients this way.
Topaz Photo AI is a great product EXCEPT your frankly offensive constant demand for obtaining our work for your AI efforts. If you want my photography, then shouldn’t you pay for it? I’ll be happy to license it to you for a very unreasonable cost, with a one-time use that will cost you to upgrade every year. If people want to give their work away to you for your development, that’s great. Let them. But make it a ONE time request, not something that we face every time we open the product. Again, this rant is not to detract from the product itself–it’s great. It’s the corporate attitude that is unwarranted.
I shoot primarily nature and wildlife (still) photography. I keep trying to use Photo AI to enhance some of my not-quite-clear photos, only to end up discarding the results in frustration because they look so artificial. In fact, I did not renew my subscription for this year because of this. I wish there were some way I could upload some of these examples for your experts to examine, so they can tell me if it’s something I’m doing wrong - or use them to figure out how to enhance your products.
My other frustration is speed. From a day of shooting, I may have 50-100 pics that I want to see if Photo AI improves them. I need to be able to import all of them at once, and quickly scan through them to identify the 4 or 5 that are worth keeping in your enhanced version. Ideally, that process shouldn’t take me more than an hour. Consider that your goal!
The last enhancement I want is a way to manually identify what I want to be the primary focal point (e.g., the eye of a bird that’s partly hidden behind branches).
I have processed well over a hundred hours of SD video with VAI, bringing it all up to amazingly clear and low-noise HD. I’ve been in the music industry in pro video editing and tape and disc replication for many years and never imagined wayback in the 80s that it would be possible to do what VAI now makes possible!
However, it does not do well with scanned 8mm and Super8. I have to use s/w to pre-clean and s/w to pre-process which makes the jobs very long before attempting anything with VAI.
My wish is that VAI can get the ability to improve and manipulate poor quality 8mm & Super8 scans that have focus, color, scratch and mold issues. A purpose-built model would help, but it would need controls specifically for color balance, saturation, contrast and gamma at a minimum. I’d also like to see an AI model like Gigapixels “Art/CG” added to VAI to ‘simplify’ detail and color range.
I agree! My biggest struggle with Video AI (and the reason I didn’t renew the product) is it’s difficult to know how to make the best choices to find good results. Templates, or analysis of the source video to create recommendations would be very helpful.
+100 on this. The product is quite expensive for hobbyists, but the results were so impressive I bit the bullet and paid $279 for the Everything Bundle. But a year later, the upgrade pricing was so ridiculous, I only upgraded Photo AI (which I appreciate the sales team offering me a decent price on). I get the constant upgrades mean developers need to get paid, but I’d like to see better pricing for existing customers. I’m happy to support great developers making great apps, but there’s a limit for us non-pros.
And the overall marketing is bordering on dishonest when it comes to pricing and sales; all these Facebook ads showing a “discounted price” that’s the same price as it was before Black Friday.
The upgrade pricing, and the advertised “sale” price, adds up to not much respect for existing customers. That’s what it looks like from the outside at least.
There are a lot of great things I like about Photo AI and Video AI, but for us M-Series Mac users, I’d love to see more focus on wringing out more performance. I know they’ll likely never be as fast as a Windows machine with an RTX 4090, but I don’t know if any optimizations have been done for M-series silicone. To be, real speed improvements would keep me renewing.
