Use speech input

The Microphone button in the Playground transcribes speech into the chat input. It doesn't provide spoken responses or full voice interaction.

note

The previous Playground Voice Mode was deprecated in Langflow 1.10. Its legacy WebSocket endpoints remain available for existing integrations.

Transcribe speech in the Playground

Speech input requires a supported browser and permission to use your microphone. It doesn't require a separate OpenAI or ElevenLabs API key.

1. Open a flow that contains a Chat Input component.
2. Click Playground.
3. Click Voice input.
4. If prompted, allow microphone access.
5. Speak, and then click Stop recording.
6. Review the transcribed text in the chat input, and then send the message.

If the microphone button is disabled, your browser doesn't support the required speech-recognition API.

Legacy voice WebSocket endpoints

The deprecated /api/v1/voice/ws/flow_tts/$FLOW_ID and /api/v1/voice/ws/flow_tts/$FLOW_ID/$SESSION_ID endpoints remain available for existing applications. They use OpenAI Realtime voice transcription, run the specified flow, and return text-to-speech audio. They require Langflow authentication and an OpenAI API key configured for the user; ElevenLabs is optional.

Don't use these deprecated endpoints for new integrations.

See also

• Test flows in the Playground
• Voice input troubleshooting