Adding the sharpening module of Photo AI 2.3.2 to Gigapixesl

I don’t know if it’s the same team working on Gigapixel AI and Photo AI, but Gigapixel, with the last version, is exactly what I need.

So first, kudos to the person(s) who are developing this software.

I use it all the time now, mostly to upscale thousands of 512x512 images. Sometimes with face recovery. Now we can load thousands of images at once without any problem.

The only thing I would like to see is a sharpening module. It’s totally fine for me right now, but I wouldn’t say no to that.

But an important note, I would like to see Topaz keep Gigapixel AI as a fast, powerful “general” tool for upscaling and cleaning up images. I still like Photo AI, I reverted to 2.3.2 because that’s the last version doing what it’s supposed to do.

So please, don’t mess Gigapixel, it’s perfect. And I hope the people working on it will keep it that way while making it work even faster.

I really want to have a workhorse that can work effectively with one image, or thousands of images.
Cleaning (noise, compression, etc) and upscale. That’s it.

Thank you for the kind words Richard!

User outreach like this is exactly what we hope for as we look to improve upon Gigapixel. Would you mind expanding upon some features you’ like to see in your ideal “general” tool?

It’s hard to say what new features should be added since Gigapixel already does a lot for me, but I think it’s easier to tell how I generally use the program, so maybe it can give a broader perspective and give some ideas:

  • 90% of my general use of Gigapixel is to upscale and clean up the compression of low- to very-low-resolution images. Coming from the web or from old pictures saved at a time when 640x480 was the norm. I use it to upscale old assets a lot. Especially transparent PNGs. So it can be used again in HD/UHD still images or videos and not be out of place with good quality elements added to the scenes.

  • I very often use Gigapixel to clean up and upscale small elements from 1080p/4k sources. Mostly low-quality sources. Blurry, with compression artifacts, etc. When I really want to restore a face in a video, for example, First, I do a general pass with Video AI to have the whole frame cleaned up without any compression artifacts. Then, to get a really clean and sharp face, I export thousands of 512x512 of the cropped face.

1.5x upscale + face recovery + minor deblur (that’s where the sharpening of Photo AI would be nice here), then composited back in the video at its original size. The 1.5x upscale + sizing down again makes the result even better. But sometimes I use 1x for exporting speed purpose.

Sometimes, I use a tool to create 10 by 10 512x512 images, so that’s 5120x5120 images to speed up the process a little bit. But I don’t upscale more than 1x and just use the face recovery, denoising, etc.

So working with relatively large images containing 100 to 200 faces sometimes is something that can slow down Gigapixel a lot (but it’s still way faster than before, so it’s still good).

  • I use Gigapixel for set extensions too. Again, I often have to work with low-resolution backgrounds. So I do all the compositing with low-resolution images in Photoshop, then I use Gigapixel to upscale and clean the result. And many times it even helps to clean up the edges of layers added in the original images.

  1. Since sometimes I have 2k to 5k images in the list, I would like to have a button to switch to a thumbnail view, with a slider to change the size of the thumbnail. Of course, in this view mode, it would remove the name and all the information about the images, but it would help a lot to find a particular image in the stack.

  2. I would like to be able to click-drag to select rows/columns that way. I do that all the time in Windows explorer to delete every 2 images of a list of thousands of files, for example.

  3. I would like to see a new model dedicated to text and/or very simple shapes with flat colors. Or having the lines, or maybe art or CG models having something dedicated to that.

  4. I have some friends who use this all the time to upscale screenshots, coming from the web or their own documents. They use it in their presentations or . It’s more professional to have something clean than a pixelated mess lol. - But this would be cool for people who edit videos on youtube. Having charts, etc, upscaled for 4k videos.

  5. I don’t know if this is possible, but I would like to see more parallelization during the export.

  6. Instead of exporting one image after the other, if the computing power is there and if it makes sense, having multiple images be analyzed and exported at once would be cool.

As you can see, I’m mostly looking for the most efficient tool possible. That can work with one image or 10k images. Simple, fast UI, no fancy animation here and there. Something that can chew everything without choking.

The fact that now I can throw in thousands of images and it loads them in 1 sec is fantastic!
Before that, I remember having to force close the program after the export, because it took so long.

1 Like

Thank you so much for the feedback, this is exactly the kind of user information that like gold to us at Topaz Labs!

I am particularly interested in your idea to allow multiple images to process at once, given enough computational resources are available.

With that said, I think you’ll love what we have in store for version 7.1.0 and beyond!

That’s cool to hear, thanks :slight_smile:
Can’t wait for V7.1 then ahah