Topaz Photo AI v2.2.2

Not sure what the heck you’re talking about wrt my post. But I am sooooooo happy for you that you are entertaining yourself, Philip! :kissing_heart:

1 Like

The product is good, but all the updates from version 1 to 2 were just buttons that were changed and that’s all - I didn’t notice any improvements in quality, the color of the images is distorted, etc. now again 99 usd for the same thing you need to pay if you want updates (in buttons), I’m not very happy, thanks, I’ll move on, I won’t update.

1 Like

How well does Photo AI run on the Ryzen 7 7840U? It has the Radeon 780M iGPU and Ryzen AI processor. How is the speed?

Topaz Labs Optimizes Photo and Video AI Software with AMD Processors and AMD Ryzen AI

https://youtu.be/IudlEiChedc?si=QTIGJpBw_3kvfBt1

When TPAI was introduced to replace Sharpen, Denoise and Giga AI, the main advertised objective was the strength of its autopilot. Years later, the autopilot and TAPI cannot be trusted. I can never ever run a batch and have confidence that results are artifact free.

Here is example, autopilot settings:

initial image:

and this is Auto Settings in Sharpen AI:

Setting to Motion correction also does not introduce the artifact in the water under the wing:

Please fix this, please give us functional software that can be trusted.

7 Likes

Error loading model - just updated 2.2.1 to 2.2.2 on latest MacOS, M1

This dialog has a bug - it cannot be scrolled to the bottom to actually read this disturbingly constantly and annoyingly appearing message. It is impossible to determine whether that is toggled on or off -
I expressly DO NOT ALLOW TO EVER UPLOAD MY IMAGES ! Please remove this “feature” - you can park it into settings without constantly nagging for this. Why, why this world is getting so entitled to use images wherever they can be gabbed. This is NOT OK.

Please STOP!

This is a commercial project not open source. Topaz Labs should hire photographers and shoot the photos away. Stop nagging for to get free photos from users who are beta testing this paid software for years.

13 Likes

The point is to see what problems you have, but they can only do that if you send pictures.

If it were up to me, many models would not exist in the software.

I use prime lenses that have few aberrations, a camera that has little noise even in the dark and almost always focuses, even when I can no longer see anything.

(The Canon R “L” lenses are really cool, and really expensive.)

I am a professional, everything that is red, in the image below, belongs away in my opinion.

Because I don’t have these problems, but the world is not that simple.



I always find it absurd that photographers often don’t give away their raw images,
they don’t have to be all you have.
Let them see what the problem is, what problems you have.
I recently asked someone if he could send me a raw, and he immediately broke off contact.
But then unfortunately I can’t help him.

I also sent TL the picture with the butterfly, the finished one and the raw one.



@ida.topazlabs

Actually, the dialog box should say that they want to see the pictures because they want to know what the problem is.

4 Likes

Correct, the message is too open ended it does not specify incidental inquiry for the image to troubleshoot a specific user issue.

I see it as a broad request for input on training data.

Again - this is a commercial project - TopazLabs is supposed to pay for this.

UPDATE: This is not an issue, a user error.

When input image is a raw file, and the only enabled adjustment is Raw Normal (e.g. strength 77, minor deblur 38) there is a discrepancy in the amount of denoising and sharpening, depending on the output:

  1. When saving as JPG, output matches the preview in TPAI.
  2. When saving as DNG: excessive blurring - both luminosity and color. deblurring is not applied

Original NEF raw file:

NEF->TPAI->JPG:


This is a good quality result

NEF->TPAI->DNG->Camera Raw->JPG:


this is significantly inferior to the above

Here is a feature request:
When processing RAW images in TAPI, allow user to adjust white and back levels to retain the full dynamic range of the image.

I think the blurring is the standard setting of camera raw.

Check the denoising setting of camera raw (color and luma).

I mostly agree with your red X’d out models.

But I have a mix of lenses. Some “pro” some not. And, mostly zoom (for what I shoot). And, I often make composites & use a mix of my own & others’ stock images. They need normalizing to work together…

Every once in a rare while the Low Res model helps an image I’m working with better than the other models. Even less frequently, the Motion Blur model helps (it mostly ruins what I work with). The Hi Def models do nothing beneficial or detrimental in my tests of them. They take up settings panel real-estate but provide no discernible value add to me (I’d expect photo realistic sharpening that is observable & I mostly don’t see any additional controlled sharpening with or b/wn those 2 models).

Thank you for pointing that, it actually is correct.
This is not an issue, a user error.

Years of using Lightroom and Camera Raw have burned their “look” into my brain.

That’s why I can recognize when they create problems.



Even when you buy presets (for Lightroom or Capture One), you quickly get one that tends to destroy images, as these are often not created by people who work with the latest technologies.

Or who make sure that the preset handles the image properly.

My comment was just an example.
No suggestion or how it should be.

I really like using the Hifi model because it enlarges the images but otherwise doesn’t affect them very much.

1 Like

Ah, that makes sense. But still would be nice if there’d be some controllable sharpening that maintains a photographic appearance.

Of all things, I just tried a surfing shot of mine from Santa Cruz, CA, and the best result (to my taste…) was the Motion Blur model (as I said, it comes in handy when it comes in handy! Though not a model that I need/use regularly).

It’s a shot with a lot of water (gentle waving), 2 surfers (going out, so on their bellies vs standing) - there’s water droplets splashing around them and you can see their hair and a “critter” at upper right.

Motion Blur model sharpening crisped the surfers (including their hair) up nicely/positively, crisped the splashes in the waves (again in a way that looked like a faster shutter speed and positively) and (I looked over the whole image zoomed in) I didn’t see [A] Water that looked more like painted strokes than fluid water, [B] Blurred patches - things I most definitely would have seen in older iterations of the product. This is observable progress in the result for this type of subject matter. Back on the Ps 2024 layer (where I launched from and saved back to) the color renditions were the same when I turned the eye icon off/on and I didn’t see any notable pixel hops (but could see the subtle (still photographic) sharpening impacts.





Has anyone had any problems with the nvidia 456.33 game ready drivers?

Bug report:

The output file size is reported incorrectly on the Save dialog:

2x enlargement is larger than the original size, here it is shown as smaller.

Since the update, Topaz 2.2.2 now takes for ever to load a photo and it takes several minutes to do anything. It used to work all the time but now, not good at all.

Ah yes, the infamous Topaz “AI” that decides where is the noise to kill and where is the texture to keep.
I see these blocky areas very often.