- TEXT
- DATA
- HUMAN
- ACTION
Short answer
How can I get real-time transcription for live audio?
You use the WebSocket session tool to link a live audio stream to your AI. Your agent will then receive a continuous stream of text as the audio is processed. You'll see the following results.
Transcription outcomes
What you get from the stream
The data your agent processes once the connection is live.
TEXT
Real-time transcripts
The AI receives text chunks as they are generated from the audio. You get a running transcript of the broadcast.
DATA
Live event logs
Your agent can track the flow of conversation. This provides a record of exactly what was said and when.
HUMAN
Manual review needed
The AI handles the heavy lifting of speech-to-text. You still need to check the output for specific names or technical jargon.
ACTION
Instant summaries
Since the text arrives live, your agent can summarize segments of the broadcast while it is still happening.
The workflow
What your AI does when the audio arrives.
The AI manages the connection and the data stream.
Open the stream
Your agent initiates a connection to the live audio source using the WebSocket protocol.
init_live_sessionProcess audio
The Gladia engine converts the incoming audio packets into text strings.
init_live_sessionFeed the agent
The transcribed text is passed directly to your AI client for immediate use.
init_live_sessionHandle the output
Your agent organizes the incoming text into readable formats or summaries.
init_live_session
Try it
Copy these to start.
Use these prompts to kick off a transcription session.
Starting points
Replace the bracketed info with your specific stream details.
Gladia Connector
You're all set. Choose your MCP client and follow the setup instructions.
https://edge.vinkius.com/vk_preview_Mlof2j7bPYQVQhjc09PLZ5CRBzJpOTb01WKt4ly7/mcpClaude Desktop
Follow the steps below to connect in seconds.
- 1In Claude Desktop, open Settings → Connectors.
- 2Click “Add custom connector” and paste the connector link above as the remote MCP server URL.
- 3Click Add and start a new chat — Gladia capabilities are ready to use.
{
"mcpServers": {
"gladia-speech-ai-mcp": {
"url": "https://edge.vinkius.com/vk_preview_Mlof2j7bPYQVQhjc09PLZ5CRBzJpOTb01WKt4ly7/mcp"
}
}
}Start a live transcription session for this audio stream.
Transcribe the live broadcast from this URL.
Monitor this live audio stream and summarize the main points as they happen.
Watch this live stream and alert me if anyone mentions the project deadline.
Claude
ChatGPT
Cursor
VS Code
Windsurf
Claude Code
JetBrains
Cline
Start here
Connect Gladia once, then ask.
Click the link to authorize the connection. Your credentials stay encrypted, and you can start transcribing immediately.
Connect Gladia to your AIFAQ
How the transcription behaves
- 01
Can the AI record the audio file?
No. This MCP is for real-time text transcription via WebSocket, not for saving audio files.
- 02
Does the AI automatically fix typos in the transcript?
No. The AI provides the text as it is transcribed. You must ask your agent to clean up or correct the text after it arrives.
- 03
How does the AI get the audio?
The AI uses the init_live_session tool to establish a connection to the live stream.
- 04
Can I use this for a recorded video?
No. This specific workflow is designed for live audio streams, not for processing uploaded files.
- 05
Will the AI interrupt the stream?
No. The AI listens to the stream provided by the session; it does not control the broadcast itself.
More questions about Gladia? The Connector page answers them. See everything the Gladia Connector can do
Connect Gladia to Claude, Cursor, ChatGPT & more
