How to interrupt voice agent?

[vaskevych]

Sometimes agent is speaking too long speech and user want to interrupt it? How to do it?

[deepgram-community[bot]]

Thanks for asking your question. Please be sure to reply with as much detail as possible so the community can assist you efficiently.
_{Consider joining our Discord community for more opportunity to engage with your fellow Deepgram users. You can earn points which can be redeemed for cool stuff by being active in our communities!}

[deepgram-community[bot]]

It looks like we're missing some important information to help debug your issue. Would you mind providing us with the following details in a reply?

• The programming language you are working in (e.g. JavaScript, Python).
• A request ID that triggered your error or issue.

[vaskevych]

It is general question and not related to programming language

[deepgram-community[bot]]

👋 You can check out Inject User Message or Inject User Agent for this need depending on how you want to handle it:

• https://developers.deepgram.com/docs/voice-agent-inject-agent-message
• https://developers.deepgram.com/docs/voice-agent-inject-user-message

This message was sent by John Vajda from Deepgram, via our community automation.

[vaskevych]

Based on documentation InjectAgentMessage is not my case as if agent is speaking InjectAgentMessage will be just ignored.
But my case is:
-agent need to stop speaking as it was interrupted by user.

what about InjectUserMessage?
The InjectUserMessage message provides a way to have the agent “hear” something without the user actually speaking it. The agent will respond as if the user had spoken the message.
What is an expected behaviour of agent when server is going to send InjectUserMessage during agent speaking?

[vaskevych]

Based on documentation InjectAgentMessage is not my case as if agent is speaking InjectAgentMessage will be just ignored.
But my case is:
-agent need to stop speaking as it was interrupted by user.

what about InjectUserMessage?
The InjectUserMessage message provides a way to have the agent “hear” something without the user actually speaking it. The agent will respond as if the user had spoken the message.
What is an expected behaviour of agent when server is going to send InjectUserMessage during agent speaking?

[vaskevych]

Do I need to use local VAD and then use InjectUserMessage?

[deepgram-community[bot]]

👋

1. InjectUserMessage
This is a purely text-based trigger. You send the JSON payload:
{ "type": "InjectUserMessage", "content": "..." }
It makes the voice agent behave as if the user had spoken that text—useful for chat/test flows or scripted interactions (deepgram.com, developers.deepgram.com).
You don’t need VAD here, since it's not tied to actual audio.

2. InjectAgentMessage & VAD
If your goal is to interrupt or interject during a live voice call—for example saying “Hello, I heard that pause”—you’d use InjectAgentMessage. This must be timed during silence, which is where local VAD comes in. It ensures you only fire the agent message exactly after the user stops speaking.

Without the VAD, you risk injecting too early or too late, resulting in ignored or refused requests.

🔍 What VAD Does Here:

When you're using InjectAgentMessage (to have the agent speak or respond), you need to wait until the user finishes talking—or at least pauses—so your injected message isn’t rejected or cut off.

That’s where local VAD comes in:
• VAD listens to the microphone input (user audio stream) and:
• Detects when the user is actively speaking
• Detects pauses or when the user stops talking

Summary:
• Use InjectUserMessage when you're sending text-based inputs (no VAD needed).
• Use local VAD + InjectAgentMessage when manually timing agent responses during live voice conversations.
• If you want to simulate a user's speech via text in a voice agent flow, use InjectUserMessage. If you're interrupting or prompting in real-time, detect silence with VAD and use InjectAgentMessage.

This message was sent by John Vajda from Deepgram, via our community automation.