Sorry, yes I was confusing mp and mb.
But what I’m finding is the remove tool does work on 20mb files ok but not on larger ones, say 50mb, on my system.
I have to test that - my photos usually aren’t over 20MB in size, mostly even <10MB as I only rarely work with RAW files.
But I would guess that it’s mostly the resolution in pixels that matters for the remove tool.
Haven’t tried the remove on that many photos yet, but neither the CR3 from my 32MP (6984*4660px) EOS M6mkII that usually weigh in somewhere between 30-40 MB nor the exported TIFF’s from Paintshop Pro (between 150-200MB) seems to give any problems.
Running Windows 10 in a VM (GPU Passthrough) with 16MB RAM and a Radeon RX 6700 (10GB VRAM) - 12 threads out of my 24 available on my Ryzen 9 5900x.
Edit:
Just for the fun of it I upscaled one of my 32MP photos by 2x resulting in a 9280*13920px that I saved as a tiff (about 640MB large). I then ran it in PAI and tested some removal. It did work, but it wasn’t exactly fast.
1 Like
You guys must be using DOS sleds for computers. I just loaded a 62MB NEF file via the Plug-in Extra, and it was converted to a 272MB DNG and it processed fine on my piddly Mac M1 mini… The total “remove” time was 42seconds for a single remove
I think it was mentioned somewhere in the beginning of the thread that this new removal tool seems to be better optimized for Mac’s at the moment and that speeds may vary for just about everyone.
Me I don’t have any direct complaints about speed in a normal “use” situation (when I’m not doubling 32MP and then load it back into PAI for the removal).
When working with 150-200MB tiffs it takes about a minute (not counting masking) to get the actuall removal done, and that is in a VM. And to fbe fair - it’s not like this tool is something I’ll use on all 2-300 photos in a shoot.
Funny enough… it only seems to be taking that “about a minute” the first time I run it, and most of that time is because of loading & preparing models. If I after my first attempt click reset and try again every attempt after that first one is done in less than half the time. Maybe there’s some room for optimization there?
For me it does however seem to be using the GPU rather than the CPU, so there’s that.
2 Likes
To read some of these posts, some of these folks act like it’s the end of their world… 
1 Like
Don’t you pay for Ps? Or, do you pay the subscription annually vs monthly?
If you have the Ps Remove functionality you’re not working with an old CS or old CC release.
I see again and again that people get GPU s with small memory recommended, 8 GB is low.
But who uses just one programme at a time and closes the others?
All programmes offload memory not only to the GPU but also the RAM, when the RAM is full it goes to the SSD and yes, 2 -7 GB/s write is slower than 50GB/s write to Ram or 288 - 1000 GB/s GPU memory.
Its 2023, people should buy 12GB and above, its Nvidias fault that we have GPUs with small memory.
2 Likes
Until recently, did not really need lots of gpu vRam. It is really only Topaz and AI that is causing me issues.
Gpus are very expensive now, even 8mp you are looking at over £200.
OK is are making money from it.
I bought CS6 years ago. It has various tools for removing or covering up objects. I don’t do it very much on my images.
Looks like I will have to buy a gaming computer, not that I ever play games, just to use Topaz remove tool.
If I had paid for TPAI, I would be getting a refund now.
1 Like
Only if they’re into gaming or really heavily into video-/photo editing stuff - or just “nerds” ;-).
I just recently bought a new system with an 8GB gfx card as this is meant for medical purposes and has to drive three big high-resolution monitors but otherwise doesn’t run anything too heavy on the GPU.
And for that 8GB is more than enough but with an Intel integrated solution you get the one or other hiccup, especially when doing RDP, so it should be a somewhat performing discrete GPU.
Still the system is fast enough also for TVAI and TPAI for the occassional use.
So, see: the needs vary considerably and thus the different offerings. No need to spend 2k+ just for a gfx card for most of the users even when they do photo/video.
1 Like
80s makes my day LOL.
1 Like
You can get a Asus 3060 V2 12GB for 300€, an that one should be as fast as my Quadro RTX 5000 16GB from 2018, as you can see in my video above, you dont need to spend 2K.
Yes, because it became so fuzzy, mane-like. 
You can try DxO free for a month, that is what I’m doing. I’m quite impressed but then I only use Raw. Im using the PureRaw3, only just came out.
Going through lots of my old images, taken on 10mp cameras.
I was so excited when CS5 intro’d content aware fill!
I went to Ps World that year to learn more about Ps, b/c it was only my 2nd gen of Ps (CS4 was my 1st). Given what a ‘miracle’ content aware fill felt like, I nearly jumped out of my skin when they brought the guy who designed it on stage to introduce him. I still use it a lot. It took me years to move from CS5 to a CC version. When I bought a new PC I finally did so… 
I’m jealous you’re not paying. A graphic artist buddy of mine just switched from CS6 to CC in the past 2-3 months, so it definitely has legs.
1 Like
No worry, we are working on that to bring the best with whatever devices you got.
2 Likes
Oh, TPAI does work well here (Mac Studio M2Ultra).
Uploaded image:
DX_2250_DPXD 1.tiff
TPAI and any other AI tool dont want to remove the powerlines,
Bin.sun, Not all of us have the latest and greatest hardware, and I’m not expecting TPAI to ever work efficiently for those of us with old machines. I like what all of the Topaz products are doing, but I just have to be judicious about which of my images I process. My 8 yr old Windows laptop with integrated Intel graphics has a whopping 128 MB of Vram, which is a tiny fraction of the 8GB recommended for the new remove tool. Using the remove tool on a jpg typically takes 14 minutes or so. Processing an image with TPAI often takes 10-12 minutes. But I continue to use the products and can’t be disappointed with how they perform. I laugh a bit when I read of others slow responses which would all appear lightening fast compared to mine. So keep up the improvement process and know that some of us aren’t expecting miracles.
3 Likes
