LogoAI ForensiX

Models


Image Deepfake Detection

AI-ForensiX AI Forensix platform provides multi-modal deepfake detection across
images, videos, audio, and text.
Each modality returns structured responses that include:

  • Prediction labels
  • Confidence scores
  • Explainability (XAI) heatmaps
  • Manipulation source classification

This page documents the Image Deepfake Detection response schema.


Response Schema

ImageDeepfakeDetectionResult

FieldTypeDescription
labelstring ("real" | "fake")Classification label indicating whether the image is authentic or manipulated.
scorenumber (0.0 – 1.0)Confidence score representing the probability of the predicted class.
heatmap_urlstring (URL) - optionalURL to the explainability heatmap showing important manipulated regions.
sourcestring ("real" | "face_swap" | "face_edit" | "ai_generated")Type of manipulation or authenticity classification.

Source Classification Explanation

ValueMeaning
realImage is authentic and unaltered.
face_swapFace swap manipulation detected (e.g., DeepFake-style).
face_editFacial attribute editing / modification detected.
ai_generatedImage generated by AI (GANs, diffusion models, etc.).

Example Responses

Listing: Fake Image Detection Example

{
  "label": "fake",
  "score": 0.967,
  "heatmap_url": "https://forensiX.com/.png",
  "source": "face_swap"
}

Listing: Real Image Detection Example

{
  "label": "real",
  "score": 0.982,
  "source": "real"
}

On this page