Skip to main content
POST
/
v1
/
visemes
Process audio and stream viseme predictions
curl --request POST \
  --url https://api.mascot.bot/v1/visemes \
  --header 'Authorization: Bearer <token>' \
  --header 'Content-Type: application/json' \
  --data '
{
  "audio": "UklGRiQAAABXQVZFZm10IBAAAAABAAEARKwAAIhYAQACABAAZGF0YQAAAAA=",
  "sample_rate": 16000,
  "model": "default"
}
'
"data: '{\"visemes\":[{\"visemeId\":1,\"offset\":0.1},{\"visemeId\":2,\"offset\":0.2}],\"chunk_progress\":\"1/10\",\"chunk_duration_ms\":100}'\n"
Process base64 encoded audio and receive streaming viseme predictions using Server-Sent Events (SSE). This endpoint is ideal for real-time facial animation based on audio input.

​
Language-Specific Models

Use the optional model parameter to select a viseme model optimized for your audio’s language:
  • "default" — English (used when model is omitted)
  • "indonesian" — Bahasa Indonesia

Authorizations

​
Authorization
string
header
required

Bearer authentication header of the form Bearer <token>, where <token> is your auth token.

Body

application/json
​
audio
string
required

Base64 encoded audio data

Example:

"UklGRiQAAABXQVZFZm10IBAAAAABAAEARKwAAIhYAQACABAAZGF0YQAAAAA="

​
sample_rate
integer
default:16000

Audio sample rate in Hz

Example:

16000

​
model
enum<string>
default:default

Viseme model to use for prediction. Different models are optimized for different languages. Available models: default (English), indonesian (Bahasa Indonesia).

Available options:
default,
indonesian
Example:

"default"

Response

Successful response streams viseme predictions using Server-Sent Events. Each event contains viseme data for a processed audio chunk.

Server-Sent Events stream with the following format:

  • Each event starts with "data: " prefix followed by JSON
  • Example normal event: data: {"visemes":[...],"chunk_progress":"1/10","chunk_duration_ms":100}
  • Example error event: error: {"message":"Error message","chunk_id":5}