Tech, ExplainedEnterprise SaaS and software
Product

Emotion AI

The face reacts before the answer comes.

BEFORE ITAFTER ITEmotion AIEntropikCOVERSFacial codingEye trackingVoice analysisHeat mapsOnline research panelssuppliesthe viewersGraphic design softwaresupplies thedraftadvertisementaddsreactionsto answersSurvey softwareBEFORE ITAFTER ITEmotion AIEntropikCOVERSFacial codingEye trackingVoice analysisHeat mapsOnlineresearch panelssuppliesthe viewersGraphicdesign softwaresupplies thedraftadvertisementaddsreactionsto answersSurvey softwareBEFORE ITOnline research panels↓ supplies the viewersGraphic design software↓ supplies the draft advertisementEmotion AIEntropikFacial codingEye trackingVoice analysisHeat mapsAFTER IT↓ adds reactions to answersSurvey software
Emotion AI receives viewers from a research panel and a draft advertisement, reads each viewer's face, gaze and voice through the webcam second by second, and adds the reactions to their survey answers.
opens a page inside this one a page in the same scene a page in another scene a concept page a label only, with no page

Emotion AI is software that reads a person's face, eye movements and voice while they watch an advertisement or use a website or app, to show how they react. It is also called emotion analytics, or facial coding when it reads the face. A survey records what a person says they think. Emotion AI records what they did on the screen, second by second, often before they could put it into words.

It measures three things. Facial coding uses computer vision, software that recognises what is in an image. It finds the face in each frame of the webcam video, and tracks small movements of the mouth, eyes and eyebrows. A model trained on many labelled faces turns these movements into expressions, such as a smile, surprise or confusion. Eye tracking works out where on the screen the person is looking. The results for many people are drawn as a heat map, which shows the most-watched areas in red. Voice analysis listens to the tone and pace of speech, in a spoken interview or a recorded comment.

The test usually runs on the viewer's own laptop or phone, through the webcam. Before it starts, the viewer must agree to be recorded, and the software should explain what is kept. Under India's data protection law, a recording of a person's face is personal data, so it needs clear consent and careful storage. The law's duties on consent apply from 13 May 2027.

Say an advertising agency in Coimbatore has made a 30-second advertisement for a pressure cooker. Before it goes on television, the agency tests it with 150 viewers from an online research panel. Each viewer watches it at home, with the webcam on. The results show that smiles rise during the family meal scene, at second 12. But at second 21, when the price and the offer appear in small text, attention drops. The heat map shows that most viewers looked at the cooker, not at the price. The agency makes the price larger and moves it two seconds earlier. Afterwards, the same viewers answer a short questionnaire in survey software, so the agency can compare what they said with how they reacted.

Emotion AI has limits. An expression is not always a feeling: a person may frown because they are concentrating, not because they dislike the advertisement. People from different places also show their reactions differently. So researchers use it together with questions, not instead of them.

How is it different from sentiment analysis? Sentiment analysis reads the words people have already written, such as a post or a review, and judges whether they are pleased or angry. Emotion AI reads the face and voice while a person is watching. The draft advertisement itself is made in graphic design software, and the findings go back to the designers.

Tech, Explained is a part of ingrid.fyi. Every page explains one piece of technology in plain words.