developers · v0.1

Send a frame.
Get back a face.

REST over HTTPS. JSON in, JSON out. The same schema whether you POST a JPEG, a base64 blob, or stream a WebSocket. Sub-20-millisecond on CPU, scales linearly per core, zero per-request setup.

Auth Endpoints Code samples Blendshape catalog Live /docs ↗

01 / auth

A token. A header.

During the private beta, requests carry an Authorization: Bearer … header. The token is scoped to your project and rate-limited per-second, not per-month — Eyeconic is built for live avatars, not batch jobs.

⏵ beta · request access from the manifesto page

http · headers

POST /v1/analyze HTTP/1.1
Host: api.eyeconic.tech
Authorization: Bearer eyc_live_••••••••••••••
Content-Type: multipart/form-data; boundary=…
Accept: application/json

02 / endpoints

Three routes. One schema.

POST /v1/analyze Multipart upload. One image, one face, full analysis JSON. The path of least resistance.

POST /v1/analyze/base64 Same response, but the image is sent as a base64 string in JSON. Better for browser → API loops.

GET /health Returns {"ok": true} plus model version. Useful for load balancers and uptime monitors.

GET /docs Interactive Swagger UI. Try a request from the browser. Open ↗

03 / code

Wire it up in four lines.

shell · bash

# single-shot analysis from a JPEG on disk
curl -X POST https://api.eyeconic.tech/v1/analyze \
  -H "Authorization: Bearer $EYC_TOKEN" \
  -F "image=@face.jpg"

python · requests

import requests

resp = requests.post(
    "https://api.eyeconic.tech/v1/analyze",
    headers={"Authorization": f"Bearer {TOKEN}"},
    files={"image": open("face.jpg", "rb")},
    timeout=2,
)
data = resp.json()
print(data["eyes"]["left"]["gaze"])
# -> [-0.18, 0.04]

javascript · browser

const grab = async () => {
  const v = document.querySelector('video');
  const c = document.createElement('canvas');
  c.width = v.videoWidth; c.height = v.videoHeight;
  c.getContext('2d').drawImage(v, 0, 0);

  const r = await fetch('/v1/analyze/base64', {
    method: 'POST',
    headers: { 'Content-Type': 'application/json' },
    body: JSON.stringify({
      image_base64: c.toDataURL('image/jpeg', 0.8),
    }),
  });
  return r.json();
};

unity · C# coroutine

IEnumerator StreamFrame(byte[] jpeg) {
    var form = new WWWForm();
    form.AddBinaryData("image", jpeg, "f.jpg", "image/jpeg");

    using var req = UnityWebRequest.Post(EYC_URL, form);
    req.SetRequestHeader("Authorization", $"Bearer {token}");
    yield return req.SendWebRequest();

    var face = JsonUtility.FromJson<EyeFrame>(req.downloadHandler.text);
    ApplyToMetaHuman(face.blendshapes);
}

websocket · streaming

const ws = new WebSocket('wss://api.eyeconic.tech/v1/stream');
ws.addEventListener('open', () => {
  ws.send(JSON.stringify({ token: EYC_TOKEN, fps: 30 }));
  setInterval(() => ws.send(grabFrame()), 33);
});
ws.addEventListener('message', (e) => {
  const face = JSON.parse(e.data);
  rig.apply(face.blendshapes);
});
// rolling p50: 14.8 ms · client-perceived

04 / blendshape catalog

Fifty-two shapes.

ARKit-compatible names. If you already rig MetaHuman, VRoid, VRM, or Apple's reference faces, the keys map 1:1. No renaming.

▸ eyeBlinkLeft

▸ eyeBlinkRight

▸ eyeLookDownLeft

▸ eyeLookDownRight

▸ eyeLookInLeft

▸ eyeLookInRight

▸ eyeLookOutLeft

▸ eyeLookOutRight

▸ eyeLookUpLeft

▸ eyeLookUpRight

▸ eyeSquintLeft

▸ eyeSquintRight

▸ eyeWideLeft

▸ eyeWideRight

▸ browDownLeft

▸ browDownRight

▸ browInnerUp

▸ browOuterUpLeft

▸ browOuterUpRight

▸ cheekPuff

▸ cheekSquintLeft

▸ cheekSquintRight

▸ noseSneerLeft

▸ noseSneerRight

▸ jawForward

▸ jawLeft

▸ jawRight

▸ jawOpen

▸ mouthClose

▸ mouthFunnel

▸ mouthPucker

▸ mouthLeft

▸ mouthRight

▸ mouthSmileLeft

▸ mouthSmileRight

▸ mouthFrownLeft

▸ mouthFrownRight

▸ mouthDimpleLeft

▸ mouthDimpleRight

▸ mouthStretchLeft

▸ mouthStretchRight

▸ mouthRollLower

▸ mouthRollUpper

▸ mouthShrugLower

▸ mouthShrugUpper

▸ mouthPressLeft

▸ mouthPressRight

▸ mouthLowerDownLeft

▸ mouthLowerDownRight

▸ mouthUpperUpLeft

▸ mouthUpperUpRight

▸ tongueOut

Got a token?

Press a button, you'll be looking at JSON in under two minutes.