Photo AI does a great job cleaning out noise in a facial shot but it seems to leave necks and torsos untouched. Is there a basic way to clean up bodies and faces at once in Photo AI? I can’t seem to find a way.
Same issue here with GigapixelAI. Their Tech does a really good job on faces, but faces end up standing out because connected (or close) skin looks so bad in comparison. I rarely can use what would otherwise be a “must use” feature for this reason.
I really hope that Topaz can come up with a way to add more skin to this algorithm whether by masking or picking similar tones.
3 Likes
I’m still looking for a solution. I’ve been trying to turn off Face recognition and have Topaz Photo look at everything as a landscape. Not much luck there…
1 Like
I have the same problem!!!
1 Like
We’re working on it! Thanks for all your feedback!
1 Like
Just upgraded, and I am having the same problem.
I have the same request. It’s a great recovery of the face but not usable for returning to a customer as a restored image. What i ususally do is blend the other areas manually with the smudge tool to “blend-in” the trashy/noisey areas to appear smooth. This takes a lot of time and effort, but works well if you have the patience to do it.
I would love to see a way to select an area to apply the face recovery algorithm to the selection and have the same great results.
I’ve been using Topaz Photo AI for a while and am reasonable happy on how this is able to recover faces,
The only big minor: the fixed -non adjustable- yellow box around the face which will be recovered. If this box is slightly adjustable I would be so happy. In the result now it’s often visible (and thus obvious) that a face is recovered, as there is a visible line from where the recovered part/box end and the blurred original face starts (see screenshot).
this is an older thread, and Tim Ho says ‘we’re working on that’ my question is - does Gigapixel8 offer this yet - your adverts seem to once again focus on face. Frankly the earlier products are amazing, but when the neck is 10% of the Facial improment then there is little point in using the product as amplifies the differences. Anyone had experience testing Gigapixel8 in this way yet ? Or does Topaz want to comment? Thanks for all the work yuo’ve done to date folks.
In Gigapixel8, it does a great job at restoring the face, but the body is pretty meh, in my experience. Nothing close to the preview images on the website, anyway
My only complaint is that the Face Recovery feature is so great that it makes all the rest of the image look Not-Good-At-All. It’d be helpful to be able to enhance the whole image.
3 Likes
can we have the option to have control over the yellow box which encompasses the face? I am presently working on images shot at 16,000 ISO which benefit from face recovery but the program often doesnt select their whole face! or, leaves out the neck.
I totally agree with the previous opinion. For me setting the fece recovery to about the half (around 40) seems to help in not creating such a big difference between face and neck, although the face recovery in the face area is then not as good as it would be with the full suggested power. Having some kind of face between the areas could already help a bit.
I agree wholeheartedly with the other commenters. I would like to be able to apply the Recover Faces algorythm to the full picture instead of just the selected face(s). For example, Photo AI does a fantastic job of getting rid of image artefacts from scanned newspaper photographs. The face comes out perfect but all the rest of the image shows the original black dots from the printing process. Please please add this option as the other enhancements (I tried them all) don’t come close to achieving the same result!
To better illustrate the comment I made previously, here’s an example of the fantastic Recover Faces enhancement applied to the original scanned newspaper photograph of my cousin Martin. The result is perfect and removes all artifacts but adding other enhancements like denoise, etc achieve nothing like this for the rest of the picture. Here it is:
1 Like
Topaz AI wasn’t designed to work on scanned images least of all newsprint scans but with a bit of preparation and as I have found imagination, good results can be attained.
I can give you the details of the work carried out if you wish.
Regards Bob.
Good evening Bob,
What you did with the picture looks like magic to me! I would love to know how you managed to clean up the rest of the picture. I did try many variations of the other enhancements but could not achieve anything as good as what you’ve done here. Great job! Was this all within Photo AI or did you use other software?
I did not try the other software I own like Luminar Neo Pro or On1 Photo Raw on this picture but I was planning to. Maybe I wont need to if results that good can be attained within Photo AI.
Thanks! Dan
Hello Dan. Nice to hear from you.
Its gone midnight where I am and I need some sleep but I will get back to this tomorrow. I checked your profile and you have the software needed just to make a slight change to your picture before you open it in Topaz Photo AI.
As Arnie said, “I’ll be back!”
Goodnight.
Good evening Dan. Sorry I’m running a bit late. I tried to post but something messed up. I’ll try again tomorrow.
Regards Bob.
Good evening Dan. Sorry I’m a bit late getting back to you.
Here as promised are the instructions on how to do the work on your picture.
First though let me give you my system details. I’m on Windows 11 Laptop, using Topaz Photo AI 3.4.3 in standalone mode.
Work to be carried out: Take a scanned image of a very, very noisy picture from a Newsprint item, preparing it briefly in Affinity Photo 2 and then using Topaz Photo AI to render a decent quality picture.
Take your original picture and open it in Affinity Photo 2.
Select FILTERS, BLUR, GAUSSIAN BLUR and move the RADIUS to 0.5 or 1 and APPLY it. Next save the picture as a COPY and close the program. Note GAUSSIAN BLUR removes some of the edginess from a Newsprint picture and softens it depending on how much the Radius is applied. EXPERIMENT WITH IT.
Open the copied picture in Photo AI. Close all AI filters that have automatically opened in the ADD ENHANCMENT stack on the right except for RECOVER FACES. Add the DENOISE AI model to the stack and under AUTO SELECTION there is a downward pointing arrow click on that to open the pulldown menu and select CUSTOM. Now select EDIT SELECTION (Above) This is where you paint in a mask which will control the area where denoising will take place. At the bottom of the DENOISE module a BRUSH SIZE and below that PAINT/ERASE. Select the left hand button (PAINT) and paint in a mask in the same area that I have in the FIRST MASK picture I have posted below. You use the /ERASE to trim your mask should you make a mistake. You don’t need to be ultra precise when using the brush as it is feathered
Next you need to go to the top of the DENOISE module and Select Controls. Select NORMAL, Strength 50, Mild Deblur 100, Original Detail 0. Note: The Mild Deblur 100 just gives the shirt a little extra definition.
Next you need to add a second DENOISE Model to the stack and Select STRONG, Strength 30, Mild Deblur 1, Original Detail 0. And that’s it.
If you want to UPSCALE the picture x2 then use STANDARD v2 with all the settings set to 1.
I don’t know how it will look if you want to upscale further, I have not tried it.
AS mentioned previously Topaz Photo AI isn’t designed to work on these types of pictures but with a little preparation good results can be attained.
I hope this helps you. Regards Bob.
Note : The third picture is rather impressive. Well done TOPAZ staff 
PS. I suggest you download the information and pictures I’ve posted as it might be deleted or moved because this part of the Forum is for ideas. Ooops! 
2 Likes
