Why use it?

  • FaceReader is the complete facial expression recognition software.
  • Objective and unobtrusive observations.
  • Saving valuable time and resources.
  • Accurate modeling of the face using 500 key points.
  • Easy integration with eye tracking data and physiology data.

How does it work?

FaceReader automatically analyzes 6 basic facial expressions, neutral, contempt, and boredom, interest, and confusion. It also calculates gaze direction, head orientation, and person characteristics. Plus 20 Action Units. The Project Analysis Module is ideal for advanced analysis and reporting: you quickly gain insight into the effects of different stimuli.

Who uses it?

FaceReader is used worldwide at more than 900 universities, research institutes, and companies in many markets, such as consumer behavior research, usability studies, psychology, educational research, and market research.

  • Market research: How do people respond to a commercial’s new design?
  • And much more!

What you should know about FaceReader, our facial expression recognition tool

Facial expression recognition accuracy

According to a recent validation study, FaceReader 6 shows the best performance out of the major software tools for emotion classification currently available, with an average of 88% [1]. FaceReader 7.1 achieves an even higher score, with 93% [2].

FaceReader Sites

FaceReader is used worldwide at more than 900 universities (including 6 out of 8 Ivy League universities), research institutes, and companies in many markets such as psychology, consumer research, user experience, human factors, and neuromarketing.

Facial expression recognition publications

FaceReader is by far the most cited facial expression recognition software in more than 1000 peer-reviewed publications since 2005 [3].

  1. Stöckli, S.; Schulte-Mecklenbeck, M.; Borer, S. & Samson, A.C. (2018) Facial expression analysis with AFFDEX and FACET: A validation study. Behavior Research Methods, 50 (4), 1446-1460.
  2. The average performance of FaceReader 7.1 was measured with the datasets ADFES and WSEFEP.
  3. Google Scholar