​
Overview

Simli Auto is a powerful API that enables you to create and manage interactive AI avatars with real-time text-to-speech and facial animation capabilities. It provides a streamlined way to integrate conversational AI agents into your applications with minimal setup.

​
Key Features

  • End-to-End Sessions: Create interactive sessions with AI avatars that can engage in real-time conversations
  • Custom LLM Support: Bring your own Language Model by providing an OpenAI-compatible API endpoint
  • Multiple TTS Providers: Choose from various text-to-speech providers including ElevenLabs, PlayHT, and Cartesia
  • Multi-Language Support: Support for multiple languages through Deepgram’s language models
  • Session Management: Built-in session handling with configurable idle times and session lengths
  • Transcript Generation: Optional conversation transcript generation for record-keeping

​
Available APIs

​
Session Management

  • POST /createE2ESessionToken: Creates a new session token for authentication. Requires your Simli API key and optionally a TTS API key.
  • POST /startE2ESession: Initializes a new end-to-end interactive session with an AI avatar. Configure TTS provider, language, and other session parameters.
  • GET /session/{agent_id}: Retrieves session information for a specific agent using header-based authentication.
  • GET /session/{agent_id}/{session_token}: Retrieves session information for a specific agent using URL-based token authentication.

​
Conversation Management

  • GET /getE2ETranscript/{sessionId}: Retrieves the conversation transcript for a specific session if transcript generation was enabled.

​
Custom Integration

  • POST /textToVideoStream: Converts text to video stream with facial animations and speech.
  • POST /audioToVideoStream: Converts audio input to video stream with facial animations.
  • POST /getIceServers: Retrieves ICE (Interactive Connectivity Establishment) servers for WebRTC connections.

​
Getting Started

To start using the Simli Auto API, you’ll need:

  1. A Simli API key
  2. (Optional) API keys for your preferred TTS provider
  3. (Optional) Custom LLM configuration if not using the default model

​
Quick Start

Here’s a basic flow to get started:

  1. Create a session token:
POST /createE2ESessionToken
{
  "simliAPIKey": "your-api-key",
  "ttsAPIKey": "your-tts-api-key"  // Optional
}
  1. Start an end-to-end session:
POST /startE2ESession
{
  "apiKey": "your-api-key",
  "faceId": "your-face-id",
  "ttsProvider": "Cartesia",  // Or "ElevenLabs"
  "language": "en",
  "createTranscript": false
}

​
Using Custom LLMs

Simli Auto supports integration with any OpenAI-compatible LLM API. To use your own LLM:

{
  "customLLMConfig": {
    "model": "your-model-name",
    "baseURL": "https://your-llm-api-endpoint.com",
    "llmAPIKey": "your-llm-api-key"
  }
}

​
Session Configuration

You can customize various session parameters:

  • maxSessionLength: Maximum duration of the session in seconds (default: 3600)
  • maxIdleTime: Maximum idle time before session timeout in seconds (default: 300)
  • systemPrompt: Custom prompt to define the AI’s behavior
  • firstMessage: Initial message from the AI when session starts

​
Best Practices

  1. Session Management

    • Keep track of session tokens and handle expiration appropriately
    • Configure idle times based on your use case

  2. Resource Optimization

    • Use appropriate batch sizes for audio processing
    • Handle session cleanup when no longer needed

  3. Error Handling

    • Implement proper error handling for API responses
    • Monitor session status and handle timeouts gracefully

​
Rate Limits and Quotas

Please refer to your API plan for specific rate limits and quotas. Ensure your application handles rate limiting appropriately to maintain optimal performance.

​
Next Steps

  • Explore the detailed API reference for each endpoint
  • Check out our example implementations
  • Join our community for support and updates

For more detailed information about specific endpoints and features, navigate through the API reference sections.