Skip to content

Use case / Voice Agents & Digital Humans

Conversational AI

Connect speech, models, tools, and devices around a live voice conversation.

Conversational AI experience

Conversation exposes pipeline delay

A person pauses, changes direction, or speaks over the response. The agent must manage that turn while speech recognition, a model, tools, and voice synthesis do their work.

  • Latency crosses several systems

    Audio travels through recognition, model inference, tool calls, and speech output before the person hears a response.

  • People do not speak in turns

    Users interrupt, correct themselves, and expect the agent to stop or continue at the right time.

  • Production needs more than a voice

    Teams need session lifecycle, transcripts, state, errors, and a path back to a human or application.

Treat voice as a live system

Build the agent around the conversation rather than a chain of disconnected API calls. Connect your model and tools while Agora manages the real-time media path.

  • Customer service

    Handle routine questions, collect context, and route a complex case into your existing support workflow.

  • Learning and coaching

    Give a learner a voice-based practice partner, tutor, or guided training experience.

  • Shopping and devices

    Help a customer compare products or add a voice interface to web, mobile, IoT, and embedded hardware.

Connect the agent to a live channel

Your server starts the agent. The agent joins an Agora RTC channel, listens to the user, runs ASR, sends text to the LLM, and publishes TTS audio back.

  • Bring your model

    Connect a supported LLM or compatible custom model, along with ASR and TTS services.

  • Manage the turn

    Support interruption, agent speech control, and session state across the conversation.

  • Receive live state

    Use RTM or the data channel for transcripts, agent state, metrics, and error events when configured.

  • Deploy across surfaces

    Connect clients across web, mobile, desktop, IoT, and embedded device experiences.

Person speaks to Agent responds

The agent and client meet in an RTC channel. Your server owns credentials, agent lifecycle, model configuration, and tool access.

AI responses can be inaccurate. Give users a clear way to review important information and reach a person when the task requires human judgment.

  1. Person speaks
  2. ASR → LLM + tools → TTS
  3. Agent responds

Own the intelligence and the experience

Choose the AI stack

Pair the real-time media layer with the model, voice, knowledge, and tools your use case needs.

Design the handoff

Connect agent state and transcripts to your interface, business logic, and human escalation flow.

Reach the right device

Use one conversational architecture across screens, mobile products, and voice-enabled hardware.

Agora network

Agora powers more than 80 billion minutes of real-time voice and video each month.

Agora Conversational AI overview

Put your agent in a live conversation

Create an Agora account, then start with the official Conversational AI quickstart for your backend.