Topaz Photo AI slow

I’ve only been using Topaz Photo AI a couple of weeks, but finding it very slow, particularly after installing 1.5.3. This may simply be because my laptop isn’t powerful enough, though according to Topaz guidelines it mett the minimum requirements. My laptop details are:

|Processor|11th Gen Intel(R) Core™ i5-1135G7 @ 2.40GHz 2.42 GHz|
|Installed RAM|16.0 GB (15.7 GB usable)|
|System type|64-bit operating system, x64-based processor|
Graphics: Intel Iris Xe Graphics

It’s taking somewhere in the region of 10 minutes per image. I’m working with jpegs, mostly just using Autopilot as there’s no way I can tweak a bit and see what happens with these response times. Changing processor preferences to CPU, graphics card, or auto makes no discernable difference.

Is there any way of speeding things up? Or do I just have to wait till this laptop dies and replace it with a high-powered gaming machine to get better graphics capability?

I started with 1.5.3 and it’s super slow. My system is high end and takes minutes per image. Whole app seems to hang while going thru the images.

Please update your graphics card drivers. Then, manually select the GPU to use for processing. Go to the Topaz Photo AI Menu option > Preferences > General menu > AI Processor > Intel Iris Xe Graphics.

Well you could blow me down with a feather. My OH (who is tecchie and normally right) had insisted that my graphics card drivers were up to date, saying that they were updated automatically by Windows, but I decided to investigate myself, and my pooter installed some new files. And now, Autopilot is taking around a minute per image (done singly - I’ve not tried putting a batch in). Obviously faster would be better, but I’m guessing I’m now getting what I could expect for my hardware.

Thank you Tim - much appreciated :slight_smile:

1 Like

Sorry, I literally just built this system. Windows 11, water cooled, RTX 4090 GPU, Ryzen 7950x CPU, 96GB RAM. There is nothing that needs updating on this system. I posted this elsewhere but have noticed the way the application works, if I load 10 images it’s super fast to process each image and most of the time the GPU is in use. If I load 20,000 images it takes a minutish to process each image and a lot of time spent with no system utilization doing who knows what. Trying to figure out how to use the CLI to process one image at a time based on my findings but seem limited and can’t set the individual settings listed in the GUI.

The CLI only processes with Autopilot. 20,000 images is quite a large batch. We generally processing much smaller batches than that to improve performance, no matter the capability of the computer.

Thanks. Thinking Photo AI needs some tweaks. Probably around whatever causes this as doesn’t look like a lot of resource utilization when the images finally load, just everything is a lot slower. Don’t know that the prior apps had this kind of issue with loading a lot of images. Any recommendation of optimal batch size? Like where does the speed take a hit? I can sometimes tell when it’s faster when I see the prior two images saving and it’s already processing the third one if that makes sense.

Also is there a way to save my settings and reload them if I’m going to be loading them say 500 at a time?

Try manually selecting the GPU to use for processing. Go to the Topaz Photo AI Menu option > Preferences > General menu > AI Processor > [your GPU, it should be the option below Auto and CPU].

The optimal batch size depends on your computer. It’ll just take some experimentation to find the ideal size to balance speed and performance.

Already set. I was asking if there is a way to save my manual settings for sharpening and face recovery I set not defalting to the GPU. Few other items:

  1. When I set some manual settings I frequently forget to do the apply to all images setting and then hours go by and it used the Autopilot settings instead for them and have to go back and delete and start over. My bad but still annoying as part of the process.

  2. Is there any way to get multiple copies of the applicaiton to launch in Windows 11? Doesn’t seem like it.

  3. Is there any way to apply the settings eluded to in Autopilot or by the name I assume it does what it thinks it should do?

Thanks for clarifying. To address your points:

  1. We’re releasing more robust controls to the Autopilot settings with v2.0 on September 7. You’ll still have to apply settings to all if you manually adjust the settings but you can configure Autopilot to be more similar to what you want in the first place.
  2. Not unless you install a beta version as well as the production version. There’s no way to have multiple production copies.
  3. I’m not sure what you mean by this point. Could you clarify please?

Yeah I’d question what I wrote to. For item 3 I assume Autopilot is what it says and you can’t set things yourself? So if I am setting Sharpening to Standard, then set 80 and 1 for the two available values, plus set Subject Only to off. Then set Recover Faces to 30. Is there any way to set Autopilot so it would be locked to these settings? That way could do this thru the CLI?

The functionality to finetune the default autopilot settings is actually coming out tomorrow with v2.0! Please update when that’s out and let us know what you think!

Thanks I updated and went back to 1.5.3. Cant even convert one PNG. Didn’t even get to the convert stage, just selecting sharpening or other settings before getting there. App just closes.

I think it’d be best for you to submit a support ticket at using the beacon at the bottom right corner of the page. That way we can collect specific info about your computer and images to continue troubleshooting!

Been thru that 100 times. Get basically a wow, your system definitely isn’t lacking perrformance and dont’ see anything else.

Do you happen to be using several large display monitors at once?

It’s a remote rendering server so technically no monitor. We remote stream in when we need to.

Have you always used a remote rendering server? Which version of Photo AI are you currently using and how does it compare to v.1.5.3 and v.1.5.4?