G’MIC：寻找一条更自由的人像磨皮工作流

ENGLISH

During my Adobe years, I took some portraits. Portrait retouching often involves skin smoothing. Although I have never been especially interested in skin retouching, it is a useful skill to have, whether a commercial assignment requires it or the subject asks for it. I remember using Noise Ninja for this kind of work back then.

Recently, I wanted to pick up the skill again. Portraits are a subject we can hardly avoid in photography. But since I left the Adobe ecosystem long ago, choosing suitable software became part of the problem.

Skin retouching calls for care. I want to improve blemishes and unevenness without making people lose their natural skin texture. Pores, fine texture, and the definition of the eyes, eyebrows, and lips should be preserved as much as possible.

I had considered learning frequency separation with Affinity Photo's native filters, but it seemed complicated and never quite convincing (perhaps I simply had not made the process complicated enough :D). So I looked for another possibility: could I find a tool for skin retouching that still gave me enough control over the process?

After discussing and researching the options with AI, I came across an open-source project called G’MIC. Yes, I tend to look for open-source and free tools first. If one is available, why not give it a try?

ENGLISH

G’MIC is an open-source image-processing project that offers a large collection of algorithms and filters. It can be used from the command line and also provides a graphical interface called G’MIC-Qt.

What caught my attention was the number of filters and the range of ways to process images.

In the G’MIC-Qt 4.0.5 installation I used, there were 645 available filters across categories such as color, detail, repair, texture, and black-and-white processing.

Of course, having many filters does not mean that every one of them belongs in an everyday photography workflow.

The important thing for me was not to study every filter, but to find a few tools that could fit into the workflow I already use.

This time, I started with portrait skin retouching.

ENGLISH

At first, I tried using the G’MIC plug-in in Affinity Photo. In my tests, however, the plug-in would not open 16-bit images in Affinity. The documentation indicated that G’MIC itself supports 16-bit images, so I decided to try the standalone G’MIC-Qt and look for a workflow that suited me better.

I use a Mac, so the steps below are for macOS only (my current version is 27.0.1). If you run into network issues during installation or updates, try a different network connection or a suitable proxy.

• First, install MacPorts, a package manager similar to Homebrew. Download the installer from the MacPorts website, open it, and follow the prompts to finish installation.
• After installation, open Terminal and enter port version.
  • If it returns something like Version: 2.12.6, MacPorts is installed successfully.
  • Then enter sudo port selfupdate to update the MacPorts package index.
  • When the update finishes, enter port search gmic. The results should include something like gmic-qt.
  • If you see gmic-qt, install it with sudo port install gmic-qt. This will install many dependencies and may take a while.
  • When installation is complete, enter gmic_qt in Terminal to launch it. The command uses an underscore, not a hyphen.

ENGLISH

The filters themselves were not the only problem I needed to solve during this exploration.

I usually develop RAW files and make basic adjustments in DxO PhotoLab, then move them into Affinity Photo for further editing. The format and bit depth of the intermediate file therefore matter too.

At first, I tried importing TIFF files directly into the standalone G’MIC-Qt, but could not get it to work, either through the file chooser or by passing the file path to the program. After some trial and error, I found that PNG was a more convenient intermediate format with the standalone version I was using. PNG supports 16-bit RGB and can therefore carry high-bit-depth images, though it is still important to check the export settings and the bit depth of the processed result.

I exported the photo from Affinity Photo as a 16-bit PNG, then processed it in G’MIC-Qt.

ENGLISH

Typing “skin” into the filter search box quickly brings up one called Easy Skin Retouch.

It offers controls for basic smoothing, detail strength, and reducing redness.

Unlike a simple blur, it lets you adjust detail at different scales. That means we do not have to rely on a single strength setting to determine how smooth the entire face becomes.

For my first test, I used the default settings.

The effect was clear: details on the forehead and cheeks became softer, and some fine skin texture was smoothed out, while the eyes, eyebrows, and lips remained mostly sharp.

But the default effect will not suit every photograph.

I then began adjusting the settings one by one.

After several rounds, I arrived at a set of settings that I am temporarily happy with:

• Edge Sensitivity: 7
• Iterations: 1
• Low Bias: 0.7
• Very Fine: 0.7
• Fine 2: 0.7
• Medium 3: 0.6
• Coarse 4: 0.5
• Very Coarse 5: 0.5
• Reduce Redness: 0.5

Note: “Low Bias” is translated here literally as “低偏置”; I have not found sufficiently reliable documentation to confirm the parameter’s precise algorithmic meaning.

These are provisional settings I arrived at after several rounds of adjustment on the test photo. They are not a universal preset for every portrait. Skin, lighting, and output size can all affect the final result.

For me, the filter’s greatest value is the range of controls it provides. By comparing results, I can gradually find an appropriate look instead of relying on a single strength slider.

ENGLISH

After testing the filter, I thought of a way to work that better suits my habits.

Instead of asking G’MIC to retouch the entire photo in one go, I can let it generate a smoothed version, then return to Affinity Photo to decide exactly where the effect should appear.

The process is simple:

• Keep a copy of the original photo.
• Place the G’MIC-processed photo on a separate layer.
• Align the two photos, making sure their dimensions and positions match.
• Use a layer mask to control which parts of each photo are visible.

I prefer to put the G’MIC-processed photo on the bottom layer and the original photo above it.

Then I only need to add a white mask to the original layer and paint with a black brush over the skin areas I want to retouch. This reveals the G’MIC result below.

Black areas reveal the smoothed layer below, white areas preserve the original layer above, and gray blends the two.

The advantage is that skin smoothing and local control become two separate steps.

G’MIC generates the processed result; Affinity Photo determines where it is applied.

If the cheeks need a little softening, I can work only on the cheeks. If the eyes, eyebrows, lips, or hair need to retain their original detail, I can leave them untouched.

Even if I later find the smoothing too strong, I do not need to process the whole photo again. I can adjust the mask and layer opacity instead.

For me, this approach is more appealing than looking for a plug-in that automatically does everything.

ENGLISH

After retouching the skin, the other details in the photo still need attention.

I use a filter in Nik Collection called Nik Sharpener Output, which is designed for sharpening the final output.

So I prefer to place it at the very end of the editing workflow.

It is worth remembering that sharpening does not mean applying a stronger effect to the whole image. If the skin has already been smoothed, excessive sharpening may emphasize its texture and grain again.

My approach is to finish the skin retouching first, then check the final output and, if needed, use a mask to control where sharpening is applied.

If the photo needs to be resized, I also wait until the final dimensions are set before applying output sharpening.

In this way, smoothing manages skin texture, while sharpening serves the clarity needs of the final output. Each has a different job.

ENGLISH

After these tests, I have formed a preliminary workflow:

DxO PhotoLab → Affinity Photo → G’MIC-Qt → Affinity Photo → Nik Sharpener Output

• DxO PhotoLab handles RAW development, basic adjustments, and any necessary noise reduction.
• Affinity Photo is used to export an intermediate 16-bit PNG.
• G’MIC-Qt generates the skin-retouched result.
• Affinity Photo is then used for local control with layers, masks, and opacity.
• Nik Sharpener Output applies sharpening for the final output.

I use a 16-bit PNG to transfer the image between Affinity Photo and the standalone version of G’MIC-Qt.

Of course, not every photo needs every step. For a photo that does not need portrait retouching, my existing workflow is still sufficient.

So far, I have focused on testing Easy Skin Retouch. I have not yet systematically explored the other G’MIC filters or made a thorough comparison across different portraits and output sizes.

This is therefore an exploration that has produced a practical result, rather than a final conclusion.

ENGLISH

I enjoy this kind of exploration.

An unfamiliar open-source tool, after installation, testing, and repeated comparisons, has become a new option in my photography workflow.

G’MIC may not fully replace a dedicated portrait-retouching plug-in, nor does it need to handle every editing task. But it offers a broad range of processing tools, while Affinity Photo gives me flexible local control over their effects.

Together, they mean I no longer have to treat skin retouching as a step that must be completed all at once.

I can generate an effect first, then decide where to use it, how much to apply, and how to blend it with the original.

The tool does the processing; the photographer makes the judgment.

That is probably what I value most from this exploration.

Note: The portrait used in this article for retouching tests was sourced from Unsplash. It is included solely to demonstrate the skin-retouching tools and layer-mask workflow.