Yup, same impression here from my first and only attempt.
You can always have opinions about the quality of these things, but for anyone that gets the idea to take advantage of this function in the first place it was more than “good enough”.
Yup, same impression here from my first and only attempt.
You can always have opinions about the quality of these things, but for anyone that gets the idea to take advantage of this function in the first place it was more than “good enough”.
What is the process to treat the image with an added effect rather than replace it in PS? I’ve only been able to replace and expand.
With the photo open on a Ps layer I generated a series of luminosity masks for it (from bright to dark) using a luminosity mask action.
I pick one of the luminosity masks in the Channels panel. Then Ctrl (Cmd) click on it to create a selection.
With the selection active I typed an art style I wanted to try (gouache painting, oil painting, line drawing, etc) into the Firefly generative fill prompt box.
Then hit Generate. It gives me the usual 3 options I can select from (or try to regenerate). For some options I put them on multiple layers & mask to pick & choose best parts. Some were more subtle than I’d have expected, some more extreme. You can always reduce opacity, try blend modes, etc. It’s after that I used PAI.
OK thanks, I think I got the beginnings of it! But time for bed…
Where does it say the minimum gpu Ram is 8gb?
It was 4gb recently.
Is Topaz the most demanding photo editing software to run?
DxO does not need as much Ram and is quicker, Topaz is by far the slowest software, I use. Photoshop is far faster. Nik Efex is about 100x faster.
8GB of VRAM is the minimum recommended for the remove tool.
If you have less, the operation will likely be processed on the CPU and thus be VERY slow. Also, larger remove areas might not fit into VRAM so you’ll likely be better off using several small masks (especially if working on high resolution images).
And yes, here Topaz is more demanding than other softwares as it is doing the process locally on your PC and not remotely / “in the cloud” as e.g. Adobes solution.
This is all getting a bit crazy. My files start off around 150mp by the time they have been through the Raw conversion. So on a file that size, I can hardly remove anything and Topaz will takes minutes to run, whilst I can remove it in Photoshop in seconds. For me Topaz AI is not worth having. I have Sharpen and DeNoise and will continue to use them although if I use DxO, DeNoise could be obsolete. T Pai just does not work for me.
Here TPAI isn’t really much different in speed to Denoise or Gigapixel for those actions ![]()
And for the remove tool:
I guess it depends on if you’re willing to pay a monthly fee and expose your images to a third party / send them into Adobes or whomever’s cloud or if you want to have the process done locally.
Besides, having a decent GPU with 8+ GB of RAM isn’t so unusual if one wants to deal with that large images as you do (although I have the impression that you seem to mix up file-size and resolution there a bit).
I don’t pay any monthly fees.
I don’t work in file sizes, only mp. Are you saying the amount of megapixels don’t make any difference, it is just the original file size that matters? I’m finding the mp size does make a difference when using the remove tool.
TPai is quicker than DeNoise and Sharpen but does not work as well, but the remove tool is very slow.
150mp Raw?
Do you have a Phase One back or a Multishot Hasselblad or Fuji GX camera?
With Adobe you do. And the others don’t have a generative AI removal tool. And as you said TPAI isn’t slower for the sharpening/denoising/upscaling than the single apps.
Remove of course is and it has been explained to you why it is.
To file-sizes / resolution dependency: Yes, of course your VRAM bottleneck will kick in much earlier with such large images.
P.S.: My confusion about you eventually mixing up MP/MB came from this:
As the RAW conversion doesn’t change the resolution of the image but of course the file size…
Anyways, with such large images you might be better off using a Cloud approach (although they mostly also have a resolution limit - wasn’t that 1024x1024 pixels for the filled area with PS with anything bigger getting pixelated?)
No it is Canon 5DSR, it is 60MP when downloaded but then is around 150MP, after DxO raw conversion. Actually some of them are 180mp.
Anyway remove is not even working on a 50mp image, but it worked ok on a 20mp. The 50mp one was a 20mp but converted by DxO.
I don’t pay Abode monthly subscription, I have a disk. Im not keen on the monthly, cloud based subscriptions.
yes true.
I think I might be the only person who hasn’t tried the remove tool. I’ve just been using PAI for Denoise and Sharpening lately. But I have noticed that 2.1.0 is a lot quicker - at least with the Extreme Noise/Lens Blur/Face Recovery models I’m using on this set of dance competition photos. It now takes less than 5 seconds to update the preview, or do the final export, whereas in the past it could be from 10 seconds to over a minute.
It still takes a few seconds more to analyse the photo in the first place, and the subject masking really can’t understand the body positions that dancers can get into!
(Dell XPS17 laptop, i9-11980HK, 64GB, NVidia 3060+6GB, TIFs up to 200MB)
So a big thank you to whoever optimised the code!
But then you don’t have the Cloud based features like generative fill, either?
And why don’t you just develop the RAW image in DXO without any upscaling there?
So, the process should go more like:
RAW → developped in DXO staying at the same resolution → removing objects → upscaling/further enhancements in TPAI and/or PS
No I don’t have some features.
I don’t know about DxO, only been using it recently, can’t see anything about Upscaling. It is either dng or tiff, 16 bit tiff are even larger than dng, which I’m using.
If you’re not upscaling in DxO then you’re definitely confusing resolution (megapixels or MP) with file size (megabytes or MB). Your camera produces images with 50.6 MP (not 60MP). DxO will definitely not change that, but it may well increase the file size from 60MB to 150MB. An uncompressed TIFF from one of your images would be about 300MB.
Is DxO upscaling?
Not that I’m aware of. But I don’t have DXO atm and the last time I played around with it is quite some time ago.
Thus the whole confusion.
In fact a 150 Megapixel image would be total overkill as long as you don’t cut out a small piece of the photo and print it in A1+