Using Watermarks to Protect Published Works
When you publish your work on the internet, there is a risk of it being used without your permission. A common solution to this problem is using a watermark.
By adding watermarks like a pattern, a signature, or a text overlay such as "Sample" to your work, it becomes more difficult to use it as is. In addition, the noise pattern may affect the training of AI image generators, reducing the accuracy of the imitation or reflecting watermarks, thus providing a certain degree of proof that your image has been used without permission.
Adding a watermark to your work
In Clip Studio Paint, when you choose Export (Single Layer) or Export timelapse, you can add a watermark to your work and export it in a format that can be posted on social media, etc.
1. Select the Export menu
To export images, select the desired file format from the File Menu > Export (Single Layer). After you specify the save destination, the Export Settings dialog will be displayed.
To export a Timelapse video, select the File Menu > Timelapse > Export timelapse.
2. Choose the watermark type
At the bottom of the Export settings dialog, turn on (1) Watermark and/or (2) Noise pattern, then tap Watermark settings.
(1) Watermark: You can set any image as a watermark.
(2) Noise Pattern (Ver.3.1 and later): You can add a noise texture to your work.
3. Set a watermark
In the Watermark settings dialog box, you can adjust the intensity of the Watermark and Noise Pattern effects.
Watermark
Import any image from your Files or Photo Library and use the handles on the preview screen to specify the placement, size, angle, you can also adjust the watermark opacity and blend mode.
Noise pattern
Increasing the Intensity, by using the slider while checking the effect, will strengthen the noise.
Once you have decided on the settings, tap OK to close the Watermark settings dialog.
4. Export Images and Timelapse
Tap OK in the Export Settings dialog to export.
Once the Export preview dialog box appears, check the image and tap OK.
Adding a watermark to multiple pages (Ver. 4.0 and later) [EX]
With Clip Studio Paint EX version 4.0 and later, you can simultaneously add a watermark to multiple pages of a comic book, booklet, or other work files.
1. Select the Export menu
Select File menu > Export multiple pages > Batch export.
When the Batch export dialog box appears, select the save destination from the Export folder and the File format, then tap OK.
By checking Split spread pages, a watermark will appear on each side of the page in the preview.
2. Select the type of watermark
The Export settings dialog box will appear.
At the bottom of the Export settings dialog box, turn on Watermark and/or Noise pattern, then tap Watermark settings.
3. Set a watermark
In the Watermark settings dialog box, you can adjust the intensity of the Watermark and Noise pattern effects.
To apply the watermark to the front and back cover pages as well, check the Set as cover page/back cover page checkbox.
You can also specify the page you wish to view by selecting the page number from the Preview Page pull-down menu.
A watermark will be applied when exported.
Your settings are saved in the app, so you can apply the same settings when exporting other projects.
Measures against generative AI learning technology
Watermarks can also be used as a measure to counter machine learning for generative AI image generators.
As AI generators learn by analyzing colors, shapes, and density of an image including its watermarks, so utilizing watermarks and noise patterns, influence the output of the generated image in addition to the prompt used. Hence, this reduces the accuracy of the imitation and may lead distortions or the inclusion of artefacts from the original image.
Generative AI image technology learns from countless illustrations to identify similar styles, content, and characters, and then generates illustrations that imitate them when prompted. Training an AI image generator on images with a specific style or selected content to reproduce that content more accurately is called fine-tuning.
Introducing three types of watermarks (A: Arabesque pattern, B: Stripes, C: CLIP STUDIO PAINT logo) which are used with the following settings.
Watermark Settings
Opacity: 20%
Blending mode: Normal
Tiling: Repeat
Tiling direction: Vertical and horizontal
Illustrations used to check the effects
We prepared 15 illustrations by the same artist to fine-tune an AI image generator. These illustrations were newly created in-house and were used exclusively for this purpose.
Using the AI image generator, we generated an image using the prompt "sks style, 1girl, brown hair, masterpiece, best quality, portrait", which resulted in the following images:
Results by watermark pattern
None
A
B
C
The degree of watermarks appearing as artefacts in the generated images is dependent on the watermark settings, but they were reflected in the generated images.
Taking measures using noise patterns
If images with a Noise pattern overlay are used for fine-tuning, it is expected that the quality of images generated by the AI image generator is reduced. By setting the noise pattern intensity/visibility to a high value, you can increase the impact it has on the fine-tuning of the AI generator.
We trained this particular model by using the illustrations we used previously with the watermark, but added the Noise pattern instead.
Noise pattern settings
Noise intensity: None, 20, 40, 60
When the AI image generator was prompted to generate the image using the same terms as before, the following images were generated.
Results by noise intensity
None
20
40
60
The None settings imitate the style of the illustration used for fine-tuning, but as you increase the noise intensity, the image quality and the generated results overall decrease. The composition also becomes increasingly distorted.
The effectiveness of watermarks and noise patterns varies depending on the AI image generator and the influence of fine-tuned algorithms used to create the images. Therefore, we cannot guarantee which results will be reflected in which environment and with which Ai image generator.
Comment