<aside>

Goal:

Build a reusable AI voice assistant that can receive voice messages on Telegram, transcribe them, respond using an LLM, and reply back with natural-sounding audio using ElevenLabs.

Problem:

Most Telegram bots are text-only and lack voice interaction. There's no easy plug-and-play AI voice assistant that supports transcribing, reasoning with context, and replying with human-like speech.

Solution:

A modular and reusable workflow using n8n, OpenRouter, and ElevenLabs that:

• Receives voice/audio input from Telegram.
• Transcribes it to text.
• Sends the transcription to a chat model via OpenRouter.
• Converts the response into voice using ElevenLabs.
• Sends it back to the user as a Telegram voice message

🔹 Step 1 – Trigger: Telegram Bot Trigger (listens for new voice/audio messages).

🔹 Step 2 – Download file from Telegram using "Get file".

🔹 Step 3 – Transcribe audio to text using speech-to-text node.

🔹 Step 4 – Send transcription to AI Agent (connected to OpenRouter chat model). Process natural language reasoning via OpenRouter (e.g., GPT-4, Claude, etc). AI Agent generates a textual response.

🔹 Step 5 – Text-to-speech conversion using ElevenLabs.

🔹 Step 6 – Send voice message back to user.

🔹 Step 7 – Trigger: Webhook (POST) receives user input (e.g., from ElevenLabs Voice Agent)

🔹 Step 8 – Message is passed to a Perplexity

🔹 Step 9 – Response is routed to an AI Agent that can use memory, context, or tools if needed.

🔹 Step 10 – The final AI-generated response is sent back via Respond to Webhook.

Tools Used:

• n8n
• Telegram API
• ElevenLabs (Text to Speech)
• Perplexity
• OpenRouter (ChatGPT, Claude, Mixtral, etc.)

Impact:

• Created a fully functional voice assistant pipeline.
• Can be repurposed for AI agents in customer service, local business bots, personal assistants, or interactive Telegram experiences.
• Plug-and-play architecture. Reusable across other voice-AI projects.

🧠Learnings:

• Hands-on practice with multi-modal AI (audio-text-audio).
• Seamless orchestration using n8n.
• Better understanding of integrating AI agents with tools (RAG, speech, webhooks).
• How to build modular AI workflows that can be reused and extended easily.

</aside>

📂 Download & Explore the Workflow:

Voice-AI-Assistant-with-Telegram-ElevenLabs---n8n-/AI Agent Into a Voice Assistant.json at main · AlvLeoAI/Voice-AI-Assistant-with-Telegram-ElevenLabs---n8n-