Skip to content
DENDRITES AI

FACEACTIONS

Read the face. Quantify the feeling.

Real-time facial-expression analysis with quantified emotion, valence, and arousal — for research, UX testing, academic study, and content evaluation.

Access is gated — usage is reviewed against our Acceptable Use Policy before activation.

SESSION · #36

12.4s · 1 participant

QUALITY 92%

EMOTION DISTRIBUTION

Happy 64%
Neutral 22%
Surprised 9%
Other 5%

Valence

+0.42

Arousal

0.38

FACS · ACTION UNITS

AU01AU04AU06AU12

AI summary · Predominantly positive engagement. Participant displays moderate happiness with brief surprise around 4–6s. No negative valence detected.

Designed for ethical, consent-based use.

Use is gated by our Acceptable Use Policy. Workplace surveillance, hiring decisions, educational assessment of individuals, law-enforcement applications, and clinical-grade medical use are prohibited. We require explicit subject-consent attestation at upload. Read the policy →

WHAT IT MEASURES

Quantified emotion. Per frame.

Eight emotion classes, valence, arousal, head pose, gaze, facial Action Units (FACS), and LLM-narrated summaries — all exportable.

8 emotion classes

Neutral · Happy · Sad · Angry · Surprised · Scared · Disgusted · Contempt

Valence + arousal

Continuous affect dimensions, frame-level, exportable

Head pose & gaze

Pitch / yaw / roll · eye state · gaze direction

FACS Action Units

Industry-standard facial muscle group activations

LLM narration

Natural-language session summary; per-speaker if multiple

Session-level export

CSV, JSON, video overlay — bring data into your stack

Multi-participant

Track multiple subjects in one video, separated by face ID

Quality flags

Per-frame confidence; low-confidence frames flagged

WORKFLOW

Capture. Analyze. Export.

Three concrete steps from raw video to research-ready data. No Windows install, no per-seat license, no on-prem hardware.

01

Capture

Webcam in real time, recorded video file (MP4 / MOV / WebM), or mobile browser. No installer — runs in any modern browser.

Multiple inputs

02

Analyze

Per-frame inference for 8 emotion classes, valence, arousal, head pose, gaze, and FACS Action Units. Multiple subjects in one video, separated by face ID. Quality flags per frame.

Per-frame · per-subject

03

Export

Real-time charts in the portal, CSV per frame, JSON event log, video overlay (tracked faces + labels), and an LLM-narrated session summary. Bring the data into your existing research stack.

CSV · JSON · video · LLM summary

BUILT FOR RESEARCHERS

Spend your time on the analysis while we handle the tooling.

Capture in the browser, get quantified emotion, valence, arousal, and FACS Action Units per frame, and export research-ready data. FaceActions handles the instrumentation so your team can focus on the questions that matter.

A UX researcher reviewing video footage on a monitor and taking handwritten notes in a calm, quiet studio.
Research, quantified

REAL FRAMES · REAL NUMBERS

Quantified reaction, frame by frame.

Tenant redacted. Subject consented to public display. This is what the portal actually looks like during a session — no rendered marketing mock-ups.

FaceActions session view — video player on the left, Expression Intensity Timeline chart on the right with eight emotion classes plotted across 81.6 seconds. Below: per-frame breakdown.

Session view

One screen brings the playback, the eight-emotion-class intensity timeline, and the per-frame breakdown together. Capture stats (frames captured, faces detected, elapsed time, processed count) sit above the chart. Export to CSV / JSON / video overlay from the top-right.

Per-frame analysis — three consecutive frames showing Emotions (8 classes with values), Sentiment & Head (Valence, Arousal, Head Orientation), Action Units active count, Facial States (Gaze, Mouth, Eyes, Brows), and an LLM-generated description below each frame.

Per-frame detail

Three consecutive frames, side-by-side, with the full instrumentation visible: eight emotion classes, valence, arousal, head orientation, active Action Units, and facial states (gaze · mouth · eyes · brows). Every frame is independently inspectable — there's no aggregation that hides the data underneath.

LLM-narrated frame strip — each frame card carries a short natural-language description such as 'neutral expression, slightly raised eyebrows' or 'slight smile, slight hint of sadness', written from the model's actual frame analysis.

LLM-narrated descriptions

Below every analyzed frame, a short plain-language description ("neutral expression, slightly raised eyebrows", "slight smile, slight hint of sadness") generated from the model's actual readings. Useful for skim review, qualitative annotation, and reproducibility checks — exportable alongside the numerical data.

WHY FACEACTIONS

Built for 2026 — not retrofitted from a desktop era.

Legacy emotion-AI tools were built when Windows installs and per-seat licensing were the norm. FaceActions starts from cloud-native and ethical-by-design defaults that the regulatory environment now requires.

Cloud-native, browser-only

FaceActions: Runs in any modern browser. Capture on a phone, analyze in a shared workspace, share results by link.

Typical legacy tool: Windows desktop installer per machine. Files exchanged by USB or SharePoint.

Multi-tenant workspace

FaceActions: One portal, many tenants. Lab and industry side-by-side with separated data, separate users.

Typical legacy tool: Per-seat licenses; data lives on individual researchers' laptops.

LLM-narrated session summaries

FaceActions: Plain-language narrative of what happened in the session — per subject, exportable alongside the numerical data.

Typical legacy tool: Numbers and charts only; narrative coding done manually after the fact.

Ethical-use gating, built in

FaceActions: Workplace surveillance, hiring decisions, educational assessment of individuals, law-enforcement applications, and clinical-grade medical use are prohibited at the AUP and product level. Consent attestation logged at every upload.

Typical legacy tool: Use-case enforcement left to the buyer; broad permissible use sometimes inconsistent with EU AI Act and BIPA.

EU AI Act + GDPR Art. 9 ready

FaceActions: Designed against the 2026 regulatory baseline. DPIA available on request; sub-processors named at /security.

Typical legacy tool: Retrofitting compliance into legacy architecture; DPIA often customer-authored.

Calibrated language about AI

FaceActions: We detect facial expressions associated with emotional states — not "emotions" as ground truth. Per-frame confidence reported; low-confidence frames flagged.

Typical legacy tool: Marketing language that implies direct emotion detection without qualification.

"Legacy tools" refers to the category broadly — established desktop-installed emotion-analysis software in active use across psychology, neuroscience, UX, and market research. We respect the category; we built FaceActions to give research teams a modern alternative on the same data shape.

PERMITTED USES

Productive, ethical, evidence-based.

The use cases we built for. Anything outside this list, we'll discuss before activation — and some categories (clinical, hiring, surveillance) are off-limits regardless.

UX research

Moderated and unmoderated usability studies. Quantify reactions across cohorts.

Content evaluation

Test video and ad creative against measured viewer expression and arousal.

Academic research

Reproducibility-friendly export. Academic license available for university research groups.

Accessibility

Detect attention and emotional state for adaptive interfaces and assistive tools.

Market research

Focus-group reaction data, quantified and exportable.

Training & coaching

Self-paced practice playback for public-speaking, sales coaching, and interview prep — with consent.

ETHICS & APPROACH

Emotion AI, done thoughtfully.

Emotion-recognition AI is one of the most regulated AI subcategories in 2026. We designed FaceActions to lead with the right defaults — not retrofit them under regulatory pressure.

  • EU AI Act — we do not enable workplace or educational emotion recognition.
  • GDPR Art. 9 — facial biometrics treated as special-category data; explicit consent required at upload.
  • BIPA / state biometric laws — consent flow logs subject attestation at every upload.
  • No clinical use — we don't position FaceActions for medical diagnosis or telehealth, and we don't sign BAAs. If you need clinical-grade tools, we're not the right vendor.
  • Calibrated language — we detect facial expressions associated with emotional states, not "emotions" as ground truth.

ACADEMIC & RESEARCH

Licensed for the way research actually works.

Discounted academic licensing for verified universities and research institutes. Reproducibility-friendly export (CSV, JSON, video overlay, LLM narration), cite-ready methodology page, and consent-attestation logs that satisfy IRB review and GDPR Article 9.

Request academic pricing →

Academic discount

Verified universities + research institutes

IRB-ready

Consent attestation logged per upload

Reproducible

Frame-level CSV / JSON · stable methodology

Citable

Methodology page with version + revision history

START

Ready for serious emotion research?

Talk to our research team. We'll walk through your use case, confirm fit, and get you onto a license that matches.