BeEmotion.ai
Sign inContact us
Cloud

A vision API for people in images and video.

Hosted AI inference engines, returned as structured JSON: faces, expressions, head and body pose, identities, transcripts.

Start freeRead the docsSign in

$5 of free credit for new accounts. Pay per task after that. Pricing · Live demo

Images

Faces, expressions and pose from a single image.

Send a photo with the tasks you want and get every face back with its box, landmarks, expression scores and head pose, plus body keypoints and identity matches when you ask for them.

terminal
curl https://api.beemotion.app/v1/image/analyze \
  -H "Authorization: Bearer $BEEMOTION_API_KEY" \
  -F [email protected] \
  -F tasks=face,emotion
Video

Video analysis with timelines and transcripts.

Submit a video as an async job and get a signed webhook when it finishes. Faces become tracks, each with its own expression timeline, and the audio track becomes timed transcript segments.

webhook payload
{
  "object": "video.prediction",
  "status": "completed",
  "tasks": ["face", "emotion", "pose", "transcript"],
  "summary": {
    "tracks": [ { "track_id": 1, … } ],
    "transcript": {
      "language": "en",
      "segments": [
        { "start": 0.4, "end": 5.1, "text": "Okay, we'll reset…" }
      ]
    }
  },
  "usage": { "units": …, "cost_usd": … }
}
Agents

An MCP server and an agent skill.

Add the MCP server to Claude Code, Cursor or Codex with your API key and install the agent skill next to it. Agent calls are billed and logged like REST calls.

terminal
claude mcp add --transport http beemotion \
  https://api.beemotion.app/mcp \
  --header "Authorization: Bearer $BEEMOTION_API_KEY"

# Then ask your agent:
# "How many faces are in photo.jpg, and what expression is each one showing?"
Capabilities

Every task the API runs.

Eleven capabilities for people in images and video. Ask for several tasks in one call; each tile links to its page in the docs.

People

Who is in the frame, where they are looking, their expression and how they move.

Face detectionFinds every face with a box, confidence and five landmarks.Image and videoExpressionScores seven expressions per face, with the dominant one plus valence and arousal.Image and videoHead poseYaw, pitch and roll in degrees for each face, to tell where someone is looking.Image and videoBody poseWhole-body keypoints for every person: body, feet, face and hands.Image and videoFace recognitionMatches faces against identities your organisation enrols, with consent recorded for each one.Image and video
Scene

The objects around them, from a fixed set or from labels you name.

Object detectionBoxes and labels for common objects in the scene.Image and videoOpen-vocabulary detectionFinds the things you name in a phrase or label list, with no fixed class set.Image
Language

Text in the image, descriptions, transcripts and questions in plain language.

CaptionsA brief, detailed or paragraph-length description of the image.ImageOCRReads the text that appears in the image.ImageVideo transcriptsSpeech to text for a video's audio track, with timed segments.VideoChat with imagesAsk a question about an image through an OpenAI-compatible chat endpoint.Image

Recognition only matches identities you enrol, with consent. Privacy

Platform

One API for scripts, services and agents.

REST APIPOST /v1/image/analyze with a multipart file or a JSON image_url, and async video jobs with webhooks.Python SDK and CLISync and async clients plus the beemotion command line.MCP and agent skillAn MCP server and a SKILL.md so coding agents can call the API.Usage on every responseEach response includes usage with units and cost_usd.
Use cases

What teams build with it.

A sales conversation across a table with a laptop
Sales coaching

Find the moments a pitch lands

Sales managers review recorded pitches to find where a prospect showed interest or hesitation, with the transcript alongside, and coach to those moments.

A cinema audience laughing during a screening
Media production

Measure audience reaction scene by scene

Studios record test screenings and follow audience expression through each scene, so editors can compare reactions between cuts.

A player wearing a headset at a gaming PC
Gaming

Adapt pacing to the player

Game developers send webcam frames to the image endpoint and use the player's expression and head pose to adjust difficulty and narrative pacing.

Financial market data on a screen
Banking

Train advisers on real consultations

Financial institutions review recorded client consultations to find moments where clients showed confusion or concern, and use them in adviser training.

Travellers at automated passport gates in an airport
Security

Review checkpoint footage faster

Security teams run checkpoint footage through face detection, head pose and object detection to find the segments worth a closer look, instead of watching hours of video.

For pilot training and driver monitoring, see AeroMind and In-Cabin sensing.

BeEmotion models also run on edge devices, for cameras and vehicles where video should stay on site.

Explore Edge →