ESRGAN is now much easier to use. Some newer GUIs (such as Cupscale) come with Python pre-packaged and allow for easy setup. There’s a new GUI in development that’s working on adding it as well. Topaz needs to step up their model game by quite a bit.
1 Like
HitPaw, accepts only internet file types.
jpeg, png etc.
On my 17 Teraflop Radeon Pro W6800 it takes 3 seconds (2 seconds for resize) for 1% point.
So 5 minutes for one image to de-noise.
The face model did not change anything on an analog test photo.
You can’t maximize or enlarge the app window, it doesn’t accept drag and drop.
When denoising another image, there were tiles and leftover pixels that it didn’t recognize as noise.
Also all details were erased.
The app does not allow model changes with sliders, you can not set it to your own needs.
jpeg testimage: 2642 x 4177 and 5K.
Denoise Denoises 150 75Mb tiff images in the same time with this GPU.
Where it makes differences between noise and details.
So, how long will it take when the customes did not spend 1000$ to 4000$ on his GPU?
The photo apps of TopazLabs run quite snappy even on relatively small GPUs.
While the latest Gigapixel update is a big improvement over the previous versions and does many things really well, the face reconstruction is still lagging behind. Simple as that. You can complain all you want, but when you need that feature, Topaz simply cannot deliver… for now. Since all these are tools that require someone to get the job done, if the tool does not deliver, and other tool does, than you use the tool that delivers. I leave the fanboy stuff to others.
1 Like
So with my pictures all the results were bad.
I would never pay money for this software.
If anything, the software only seems to be useful for tracking people when the source material is really bad.
They also offer watermark removal software on their site.
Not sure why you have a chip on your shoulder, but that is your opinion. Personally I use the software as much as Topaz Gigapixel and it does excellent job with projects that Gigapixel cannot do. Simple as that. If you have the project that requires the right tool you use that tool. Not everyone is using upscale features for photography and only photography. People use it for all kinds of things, from up scaling logos, graphics of various kinds, art, heavily compressed images, small images, screenshots from old movies, YouTube video screenshots etc. And I personally use a lot of that for reference. And in some cases, I want a clean plastic look that I can get because it makes a better reference for other projects. So don’t use yourself as reference for what software can be used. You have your use cases and other people have theirs. The apologist attitude and bashing of competition does not help anyone. Please don’t do it.
1 Like
There is no Mac version for HitPaw yet, so I cannot comment on how good it is. Presumably it can be downloaded as a trial.
I gave MyHeritage Photo Enhancer a trial. But, however good the AI, it’s never going to reconstruct a face correctly. I tried it on old photos of me and the result, while maybe convincing to someone who doesn’t know me, looked nothing like me!
1 Like
So far I’ve seen quite a few similar Online AI photo enhancers, but HitPaw AI has three models now and one it tries to make itself different by being offline app. I guess the downside is no Mac version, but perhaps they are planing it.
So far I’ve tried every AI enlarger that is commercial, didn’t code one myself, but all the commercial options, HitPaw AI new model for reconstructing faces, is so far the best, and superior to Topaz with examples where there are small or blurry faces. In normal, high res clean images, it delivers similar results to Topaz, some better, some worse, depending on the image, but on average very comparable.
There have been threads about Tiopaz and need to add better face reconstructions of small faces, but so far Topaz have not implemented it. My point was that Topaz is releasing large applications, lots of bugs, and somewhat frustrating User interface that keeps changing and pricing structure that is not super clear and also some of their products like Mask AI is abandoned it seems, and that makes people trust the company less. By comparison HitPaw AI is maybe now 800 MB download with three good AI models to choose from. Intalation is no larger than 1 GB. By comparison with all the models and everything GigaPixel is about 6 GB. That is six times the size for same or lesser results. Not good.
The Topaz products keep getting larger in size and not proportionally improving in their ability to deliver more. There has to be more efficient way to code and deliver smaller installation than right now. I’ve complained about this and many others have for over a year now, with no change or response from Topaz.
1 Like
There are many free programs now that compare with or beat Topaz’s paid products. ESRGAN in particular has many custom trained models that are simply superior to Gigapixel and most of Topaz’s Image based products, and are free to use for consumers.
Video Enhance AI is the only product from Topaz that I haven’t seen any good competition for, paid or free. The prices Topaz are charging for inferior software or results is just astounding.
1 Like
I figured I’d give this a try as I’m very happy with Gigapixel and use it to salvage either my older digital images and film scans and to upscale smaller AI app-generated art.
First of all, HitPaw’s app doesn’t work under VM (Windows on a Mac) which you find out after you install and try to run it, and as you said there is no Mac version. So I had to fire up a PC and start again.
On the PC, I still don’t know how good the app is because it refuses to generate a preview (“0%” flickers when I hit the Preview button but nothing happens). I also can’t see the two before-and-after panels you show (I see a single panel) or export the result to view separately when using the trial version. So that’s the end of that.
And jpavatargirl, the average consumer is not going to mess with ESRGAN! “esrgan requires python >= 3.8. The requirements.txt file can be used to install the necessary packages.” Check the text file: “This page does not exist yet.” OK…
There is more and more competition and most deliver either similar or inferior results to Topaz AI programs, but some are diversifying and some are better than Topaz, because Topaz management team apparently does not care about any of the legit complaints, criticism and suggestions in this thread.
The best new company I found is called HitPaw and they make some really user friendly and powerful AI photo enhancement products. Their alternative to Gigapixel is particularly interesting and with latest update it blows Gigapixel away in terms of reconstructing faces. If Topaz does not wake up and starts actually listening to users and actually innovate again, I think people will be moving to alternatives as soon as they become available.
Here is what HitPaw AI Image Enhancer can do and its small install with offline AI models, a fraction of the size of those that we have to download with Gigapixel and its in some respects, superior to Gigapixel. Topaz, you guys better wake up.
Hmmm…
I tried the HitPaw Windows app for the Denoise model. I used one of the Topaz test images.
Topaz Denoise processed this image on my system (AMD RX570 GPU) in 5 seconds.
By this same time, HitPaw was only 3-4% completed (no GPU processing?). In other words, in this comparison the HitPaw Denoise was so slow as to be unusable. I may try some of the other functions, but I am not encouraged.
image used - Topaz test image.
If HitPaw is base on open-source AI upscale algorithm, then I think it require CUDA (Nvidia GPU) for GPU processing. 
no it was using my radeon card too.
maybe OCL.
You might have misinterpret my comment. I am not comparing Topaz GigaPixel and HitPaw as two equal programs that compete in the same way. I was commenting on the fact that until the latest release 5.8.0, was looking like they were not making many big improvements, but were more hawkish with their pricing. Lots of bugs too. Some programs pretty much abandoned, like Mask AI, Adjust AI. JPEG2RAW AI etc. Not good signs. Meanwhile there are many more companies entering the AI race and HitPaw AI was one of them. Even now it still does a better reconstruction of blurry and small faces than Topaz, and for some users its enough what it does. Point I’m making is that when companies get complacent and stop innovating , competition catches up and surpasses them.
Not sure how much you follow camera market, but remember when Nikon and Canon were kings and than Sony showed up on the market, seemingly out of nowhere and surpass them, and than by the time Canon and Nikon joined the mirrorless game, they already lost market share. Canon managed to regain a lot of it, but Nikon not so much. There are many many examples like that. I hope Topaz does not become case study because of their complacency.
1 Like
Topaz has had many updates of their AI apps over the last year; some of which changed major underlying sub-systems. I would let them worry about their business model.
It looks like HitPaw is certainly not a significant competition in the denoising arena.
On my system, the performance is terrible. On a small 400K file here are the times for the General model:
Topaz: 2 seconds
HitPaw: 2.3 minutes
If you compare it head on, I agree Topaz is much faster and for images of certian type better. You try to reconstruct small images that are heavily compressed and reconstruct faces and the story changes in favor of HitPaw. Question is what is the job you are doing and what it is the best tool for the job.
You don’t carry 400 mm f2.8 full frame lens to shoot birthday party of your neighbors sun, do you? There are better and cheaper tools for the job, and still more appropriate.
There are simply some images and needs that HitPaw does better than Topaz. Will see how it will work in the future, but for the time being some things you just can’t get with Topaz. Here is one example. By the way in case you are wondering, I’m not going for the original only larger resolution, I in face want a reconstructed face because its used as reference for other post processing tools.
There are many examples I could show, and they simply are two different programs that work differently. I use the one that I need for the specific job. Simple as that. I do not compare them for the same job, because they have different abilities. Similar to 400mm lens vs smarphone analogy.
There seems to be some kind of tribalism in this forum where anything not named Topaz is seen as some kind of threat to the holy Topaz. I do not see it as threat. I see it as another tool on the market, that if anything Topaz should emulate and add similar ability to its set of features.
It’s not really tribalism.
You have pointed out one use case where HitPaw works well.
I pointed out one use case where it fails miserably, imo.
The over-the-top chear leading started to sound a bit like an agenda. You never know these days.
But, yes, use whatever works best for you for the specific job you are doing.
Update: I did try a face reconstruction with a small (25K) jpeg and it worked quite well. However, anything larger than ~25K and it seems to slow down dramatically. I’m not clear why that would be the case, but it severely limits the use case for me.
I’m still trying to determine why the HitPaw app insists on running with Admin Privs, which is disconcerting from a security standpoint.
This test didn’t turn out so well. It looks over-processed. I’m not aware of an intensity adjustment.
In any event, I tried a few things to get the permissions back to User, but was unsuccessful.
So, unfortunately, I was left without much recourse but to uninstall.
Seems to me you chose the wrong tool for the job. The image is already processed and not the best for the thing you wanted to do. You should not blame the tool. You should be more deliberate in choosing the right tool for the job. I can only upload five mages here, but I assure you these images were not possible to save with anything Gigapixels offers at this time. But HitPaw was able to salvage the images.
Furthermore, I often use more than one AI model or even tool and combine the images for best result, on the challenging examples. If Gigapixel can reconstruct jewelry or clothing better than faces than I’ll use that for everything but face and simply compositing it later in Photoshop.
Don’t blame the tools. Become better at understanding their strengths and weakness and when and where to use which tool for best results.
Speaking of tools, adding film grain to over processed smartphone faces helps bring back the illusion of texture. Its yet another tool one can use.
oi
The original image in my post was 1024x768. So, what if I wanted to enlarge it? I can’t do that with adequate results?
I rarely have images as “bad” as the ones in your examples!
But, I agree. If you have images that “bad” then HitPaw does a good job at reconstructing them. I am not blaming the tool - It just doesn’t offer what is useful for me. If you have very bad images, as in your examples, it can work wonders to improve them.
1 Like
