Frequently Asked Questions
Here you can find answers to frequently asked questions.
What is Rewarp?
Rewarp is an App that uses on-device Machine Learning to predict and undo photo manipulation effects (commonly known as Warp, Liquify) done to a human face.
Which image manipulation effect does Rewarp recognise?
Algorithm is trained on faces warped by the Face-aware Liquify tool in Photoshop, and will thus work well for these types of images, but not necessarily for others.
- ✔️ Detect face alterations (eyes, nose, forehead, chin, cheeks, mouth)
- ✔️ work on selfies with clearly visible face
- ✔️ work well with high resolution photos (at least 400x400px for face area)
Facewarp DOES NOT:
- ❌ recognize local modifications (acne touch-up, healing brush, "photoshopped makeup" etc.)
- ❌ recognize body alterations
- ❌ work well with small photos (less than 400x400px for face area)
- ❌ work with faces other than human (animals etc.)
Are my photos safe?
Rewarp works completely offline and all the processing is done, from start to finish, directly on your device. Your photos are not being sent to server or stored anywhere.
Do I have to buy the Mobile App?
No. Rewarp is completely free to download and install. However, free version of the App has a daily limit for the number of photos you can process.
What is Rewarp for?
Rewarp is intended to be used for entertainment purposes only, we do not take any responsibility for any misuse of the results.
What am I buying?
Rewarp offers a one-time payment that removes the daily limit of photos you can process which is present in the free version.
How does it work?
A photo you select goes trough several processing steps. After you select a photo, the image data is adjusted and ran trough the two CoreML models, on your device. The resulting data then goes trough complex post processing, which generates the images you get to see as a result.
Are results reliable and accurate?
Since the ML models have been trained by a limited number of testing photos, inaccurate predictions do happen. Model is trained on faces warped by the Face-aware Liquify tool in Photoshop, and results may be biased ().
Algorithm is trained on faces warped by the Face-aware Liquify tool in Photoshop, and will thus work well for these types of images, but not necessarily for others. We call this the "dataset bias" problem. Please see this paper for more details on this issue.
What Machine Learning models does Rewarp use?
Rewarp AI models are heavily based on Detecting Photoshopped Faces by Scripting Photoshop paper.
Can we expect Android version?
Since this app heavily relies on Apple's CoreML, it's unlikely that we will publish Android version any time soon 🤷.
Need more information?
Get in touch via firstname.lastname@example.org. We will happily respond to your questions and requests.