​
Images and videos

Revelio is currently the model with the most effective architecture in many contexts. V01 and Brokenwand are on the platform for coverage reasons for specific scenarios (old generators no longer widely used).

​
V01

  • Technology: Older model along with Brokenwand. It is a model based on a technology that allows working on “general” aspects of an image, regardless of the type of image displayed. These are therefore “generalist” models and therefore allow analyzing a good number of image types but with usually lower reliability compared to models like Revelio.
  • Dataset: The artificial dataset covers the period January - December 2023.

​
Brokenwand

  • Technology: Older model along with V01. It is a model based on the same technology as V01, but with slight variations relating to the “perturbations” (compression, rotation, etc.) applied to the images used during training.
  • Dataset: The artificial dataset covers the period September 2023 - March 2024 for artificial images. The dataset used is very large, but on old-generation generators. Usually quite accurate on real images.

​
Morphing

  • Technology: This is a different technology from the previous ones, whose objective is to identify faces as a first step, and then analyze their features in search of distortions compared to the real image. It is a model that, when it correctly detects a face, allows determining if it has been transformed in some of its features. It is less effective in the case of completely generated faces; in case of no faces, this model can be ignored.
  • Dataset: Model trained on images containing “morphed” faces, i.e., modified with both classic techniques and those related to GenAI up to 2023.

​
Revelio-III

  • Technology: This is a technology that allows training a specialized model, combining the semantic part of the image with the intrinsic characteristics of the image at the pixel level. Unlike the V0X family, it does not analyze the specific detail of all pixels, preferring a “high-level” analysis combined with the semantic value extracted from the image. The advantage is greater precision on certain types of images at the expense of reliability on images not aligned with those in the training set.
  • Dataset: The data used to train this model dates back to the period September 2023 - May 2025. It should be noted that, as far as artificial images are concerned, the predominant subjects in the training set are people.

​
Ellen

  • Technology: The model’s architecture is specifically designed to detect traces left by photo editing tools, mainly in the context of image compression. It is a model that responds affirmatively only in the presence of photo editing; in other cases, it is not considered (it is listed under “excluded models”).
  • Dataset: The data used to train this model consists of pairs of real images and their counterparts modified with photo editing tools (cut&paste, splicing, etc.).

​
Audio (speech)

Phantom is currently the most effective model in most contexts. It is appropriate to take into consideration the variability and complexity of speech (types of voice, speech frequencies, etc.)

​
Zebra-II

  • Technology: the model’s architecture is structured to recognize artificial audio and real audio, based on a supplementary model (Wav2Vec2) for the numerical transformation of audios;
  • Dataset: the data used to train the model consists of real audio and artificial audio; in the case of the first version of Zebra (Zebra I), the data covers a period from December to February 2025, while for the second version (Zebra II) additional data up to May 2025 has been added.

​
Phantom:

  • Technology: the model’s architecture is different from Zebra, as no auxiliary model (Wav2Vec2 embedder) is used, but the audio converted to mono is directly used and processed;
  • Dataset: the data used to train this model consists of real audio and artificial audio, from a period ranging from March to May 2025.