6
\$\begingroup\$

(As a preamble, I am not saying anything like Topaz is dishonest, or I feel ripped off)

Watching YouTube videos I have now seen more than once seen people say they like to first use Topaz Denoise, then Gigapixel, and then Sharpen. This has been very inconsistent with my personal impressions of the 3 apps.

To me it seems like Topaz only hast one algorithm. It has two phases, denoise, and upscale. Denoise focuses o9n denoise. Sharpen denoises and then sharpens. Gigapixel denoises, changes the image size, and then sharpens. So to me it seems like an image should be fed through one of these programs, based on what is really needed.

Has Topaz ever stated if there are different AI backing these programs? Do they have recommendations on using one or all three for one image? Has anyone done tests to compare combining the programs vs using one at higher settings?

I would also like to hear if anyone has a personal workflow that they trust in, using one or multiple of the programs.

\$\endgroup\$
5
  • \$\begingroup\$ It's a bit of a wide question which is hard to answer, unless Topaz themselves pitch in. Do you maybe have a (more) specific issue that you're dealing with? \$\endgroup\$ Commented Nov 16, 2021 at 15:18
  • \$\begingroup\$ @SaaruLindestøkke if you have a good reference that Topaz recommends using multiple apps on the same image, that would be a good answer \$\endgroup\$
    – Andrey
    Commented Nov 16, 2021 at 15:27
  • 1
    \$\begingroup\$ IMHO there is no workflow that is good for all images. Each image has its own problems (possibly in light of its intended use) and the skill of the editor is to identify them and use the appropriate tools to solve or mitigate them. \$\endgroup\$
    – xenoid
    Commented Nov 16, 2021 at 16:15
  • \$\begingroup\$ @xenoid sure, but there are lots of processes that don't work. I don't think anyone would recommend using Sharpen twice \$\endgroup\$
    – Andrey
    Commented Nov 16, 2021 at 16:29
  • \$\begingroup\$ Hmm, ok, I'll see if I can find something. Could you perhaps include links to example videos you refer to? That would prevent that I (and others) would suggest resources you've found yourself already and which didn't fit your question for some reason. \$\endgroup\$ Commented Nov 16, 2021 at 17:34

3 Answers 3

8
\$\begingroup\$

You asked,

Has Topaz ever stated if there are different AI backing these programs? Do they have recommendations on using one or all three for one image?

Yes to both. Straight from the source, at Topaz Lab's support site (emphasis mine, and some text omitted):

When To Use DeNoise AI, Sharpen AI, and Gigapixel AI

One of the most common questions we receive here at Topaz Labs is when should you use each of our plugins? The truth is there isn't one "perfect way" to use our programs but a thorough understanding of each program's purpose will help you make an educated decision for your own unique workflow.

When To Use DeNoise AI

DeNoise AI is intended to be a dedicated noise reduction software. In particular, it allows you to reduce noise while retaining fine detail that would be lost in traditional noise reduction tools. While it has some sharpening capabilities, these are not the same robust AI sharpening capabilities in Sharpen AI to address concerns such as motion blur and missed focus. This is just a basic sharpening tool for when more complex sharpening is not needed, similar to what you would find in Lightroom or Photoshop.

As for when to use DeNoise AI, consider how noise is affected by your editing. Anytime you edit a photo, you alter the pixels in that photo. Your noise is found within those same pixels, and as you edit you are affecting both the image you photographed and the noise.

Most of the time, you can probably run DeNoise AI wherever you want in your workflow and it will work just fine, but in some cases, if you edit the noise too much, or if the noise is not recognized by the AI model, the noise reduction might fail. Remember, DeNoise AI's noise reduction models are trained on unedited noise.

When such a conflict is suspected or likely, you should run DeNoise AI early in the editing process before you've edited the noise itself.

[...]

When To Use Sharpen AI

Sharpen AI is intended to address several concerns including motion blur/camera shake, missed focus, and out-of-focus images. It also has some minor noise reduction capabilities similar to what you would find in Lightroom or Photoshop, but it should be noted that the noise reduction feature in Sharpen AI is not the same AI-based, detail-preserving noise reduction found in DeNoise AI. It is simply in Sharpen AI as a minor workflow tool when more robust noise reduction is not needed.

As for when to use Sharpen AI, this depends on the photo itself. Traditionally, you would do most of your sharpening toward the end of an edit, but if you're correcting missed focus or motion blur, you'll probably want to use Sharpen AI earlier in your editing process so that your subsequent edits don't interfere with the process. For more generic sharpening to address "soft" images, you will probably want to do that toward the end like you might do in other workflows.

When to Use Gigapixel AI

[...]

There are also some tools for noise reduction and blur removal, as well as color bleed reduction and face refinement built into Gigapixel as well. Like our other apps, these are not necessarily as robust as our dedicated apps for noise reduction and sharpening but can help speed up your workflow when such additional image improvement is not necessary. It should also be noted that Gigapixel has a model for animation and computer graphics images as well.

As for when to use Gigapixel, it is often used as a final workflow step but this can also depend on the original input image as well. If you want to enlarge an image you are happy with, use it last. If you need to enlarge the image in order to edit it properly, it's ok to use it first.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Thnx, that was exactly the info I was looking for. I think if there is some YouTuber not sure what to do on photography, comparing denoise from Denoise and Sharpen would make a great video \$\endgroup\$
    – Andrey
    Commented Nov 17, 2021 at 14:35
3
\$\begingroup\$

I have no affiliation with Topaz, nor is this post a recommendation. I just felt like googling and collecting below information.

Topaz has a Youtube channel with plenty of tutorials on using their tools separately or together.

A video that you might find relevant is about a Lightroom workflow where all three tools (DeNoise, Sharpe, Gigapixel) are used together on a cropped image of a bird.
The video has an accompanying article which sheds some light on how to use these tools and in what order they should be used. From the table of contents:

Step 1 – Use Adobe Lightroom Classic to correct tone, color, and composition
Step 2 – Use DeNosie AI to remove distracting noise while retaining edge detail
Step 3 – Use Sharpen AI to add sharpness and edge detail
Step 4 – Use Gigapixel AI to upscale and increase resolution without losing quality

To prevent applying noise reduction twice, the "Suppress Noise" slider is set to zero in Sharpen AI (step 3).

\$\endgroup\$
1
\$\begingroup\$

Do the 3 main Topaz apps really do different things?

Topaz released in September 2022 their new flagship program Photo AI, which combines the 3 Topaz apps you mentioned: denoising, sharpening and upscaling. All 3 steps can be done automatically:

Photo AI has a feature called Autopilot, which automatically analyzes a set of photos and applies what it determines to be the right models and settings, which a photographer can then use as a starting point to make creative edits. The idea is this platform removes the tedium of fixing common image issues manually and instead lets photographers skip right to the best part of photo editing.

\$\endgroup\$
1
  • 1
    \$\begingroup\$ Thnx for mentioning this, I was actually going to go add a note to my question about this release. I upgraded and am very happy \$\endgroup\$
    – Andrey
    Commented Sep 30, 2022 at 14:11

Not the answer you're looking for? Browse other questions tagged or ask your own question.