I’ve looked for this idea in other posts but couldn’t find any. I hope I’m not repeating an idea.
Much of my work lately involves retouching old photos for our local History Preservation Society. Many of the images began as scans from glass plate negatives. I’m dealing, usually, with soft faces and Topaz is incredibly helpful. Astounding in many cases. In the old formal images of 1880-1920, the hands are so important. So, improving the face (as Topaz does so well) only makes the hands seem worse by comparison. I know I’m asking a loaded request, but having a ‘hand recovery’ action would be tremendous. Or, even if I was able to draw a rough marque around a hand and have it be improved would be a great step forward. If such a thing is impossible, I’m still thrilled to continue with Topaz, as it’s made quite a few people in our town very happy actually being able to ‘see’ the face in focus.
I think Topaz continues to see face recovery as an appendage of Gigapixel and not as a feature that has considerable value even for images that do not require enlargement. This is the case, for example, of those who recover old photos, both b&w and color, where face recovery is essential to give value to these old images. I would suggest Topaz to invest in this aspect and to resolve/develop those requests that come from many sides: improvement of the treatment of the neck, hair, hands as suggested here, ghost glasses etc.
We hope that sooner or later they will also notice this segment of users who make face recovery their main PhotoAI requirement
When I use face recovery on PAI, on my images, it blurs the face.
It also rarely recognises there is a face, maybe because my faces, tend to be quite a large proportion of the image.
Could you include an example image?
Not sure how to do that.
I did find a previous post about it: Face Detection - Always makes faces blurry
Just working on another portrait. This face is around 10% of the image size.
On PAI, face recognition on, has blurred part of the nose and mouth area, vaguely sharpenen around the eyes, sharpened hair. With Face rec. off if does similar job, in fact the same. Also it has added some bright white artifacts to the nose, that is on both PAI versions.
Doing original image, with just SAI, does a good job, no real issues. With T DeN first before SAI, again no real issues apart from down her right arm edge, there is red/brown line, easy to remove with a layer mask.
So PAI, is making a real mess. Why is it blurring some parts anyway?
Good result, so it does work on some images.
How did it repair all the other parts away from the face?
You can select other faces on the image, not sure if you selected the hands it would work, maybe worth trying? Makes sense to be able to extend it to other areas of the image.
I don’t understand why it is blurring my faces, I may want some parts blurred but certainly not the eyes.
It is also inconsistent, blurring part of the clothes but not all, just looks weird on the image. OK I can remove that, as I use layers in Photoshop.
But it is a lot of hassle, this should be saving time not causing more work.
The remaining sections of the image was corrected by me zooming in to 200% in Photoshop and fixing all the dust particles, scratches, scuffs, etc. Very time-consuming. But, it’s the only way to remove those things without causing the underlying image to go soft further.
I will try selecting the hands and using ‘face’ correction. As you say, worth trying.
I recommend that you post some examples here so that developers can take a look at them.
But Topaz is aware of the problem.
How do you post images here?
What format and size can be used?
Just drag and drop images into the edit area or click the upload button.
It support JPG, PNG.
Size doesn’'t matter, it will auto resize if the image is too large.
Did you see the post I linked earlier? That was back in August. I’m amazed the developers were not aware of it. All very well saying it only is meant to work on low res images. Really this is not even at beta stage.
I will try to post some images later.
I am attaching a couple screen shots. This is an underwater image, I do a lot of underwater models. I run the raw image through PAI before I start any editing on it. It helps to give a cleaner bases, usually. But this face recovery has been through all iterations, unusable for me. You can clearly see the delineation between face and neck, and the ear. Also, the noise that is introduced through sharpening on the neck and skin and shadows isn’t good. If you continue with face recovery, make it skin recovery, not just one body part, Don’t limit it, but have AI find the skin. I can do that in LR and PS. Thank you. I do use PAI for everything. It seems it was better in some previous version and the more you add to it, the more noise and roughness is being introduced.
You may want to try DxO PureRaw, I find that pretty good at giving clean images, it is just a Raw processor. It also has lens correction.
I do a lot of long exposures, which tend to be noisy, it does a good job on those, at removing noise. I find Topaz blurs noise rather than remove it.
You can try it for 30 days free.
The problem with noise, if you add sharpening, it will make it worse.
Underwater is a bit tricky with the distortion and lighting.