🎭 Affective Computing — Live Multimodal Emotion Recognition

Reads emotion from your face, the tone of your voice, and the words you say — then combines them. Works on phones and laptops. Nothing is stored; everything runs on the fly.

How to use: give any one (or all) of the inputs below, then press Analyze emotion.

1) Face — take a webcam photo (or upload one)

2) Voice — record a few seconds of speech

3) Text (optional) — type/paste, or leave blank to use your speech

Combined emotion (late fusion)

Transcript (what we heard)

Notes · Facial emotion uses DeepFace · voice tone uses a scikit-learn model trained on RAVDESS (neutral / happy / sad / angry / fearful) · text uses a DistilRoBERTa emotion classifier · speech-to-text uses Google's free service (typed text is a fallback). A short, front-facing photo and a clear couple of seconds of speech give the best results.