Telegram Voice Replies with OpenAI, Gemini & ElevenLabs (n8n)

Turn Telegram voice notes into spoken AI replies—automatically

This n8n workflow listens for new Telegram voice messages, transcribes them with OpenAI, generates a conversational response with Google Gemini (with per-chat memory), converts the reply to speech using ElevenLabs, and sends the resulting voice reply back to the same chat.

What this workflow does

• Triggers on Telegram voice messages and downloads the attached audio.
• Applies your configuration (prompt, model selections, memory window, transcription language, and elevenlabs_voice_id).
• Transcribes speech to text using OpenAI speech-to-text.
• Generates a reply with a Gemini-powered agent using conversation memory scoped to the Telegram chat ID.
• Converts the text reply to audio via the ElevenLabs text-to-speech API.
• Sends the audio back to Telegram as a voice reply, with the text response included as the caption.

Use cases

• Voice-based customer support: users speak; the bot replies with a natural voice response.
• Personal or team assistant in Telegram: maintain context per chat while handling voice queries.
• Hands-free workflow triage: capture spoken questions and respond quickly with audio guidance.

Technical details

• n8n Telegram trigger to detect incoming voice messages and download audio.
• OpenAI for speech-to-text transcription.
• n8n-nodes-langchainagent (“Telegram voice agent”) to run the Google Gemini chat with per-chat memory.
• HTTP Request to call ElevenLabs TTS using xi-api-key header auth.
• Telegram send-audio action to post the generated voice reply back to the originating chat.
• Configurable items include system_prompt, memory_window, gemini_model, elevenlabs_model, and stt_language.

Setup overview

Create a Telegram bot via @BotFather, add OpenAI + Google Gemini (PaLM) credentials, configure ElevenLabs API key auth with the xi-api-key header, paste your elevenlabs_voice_id, then activate the workflow.