Turn Telegram into a smart chat assistant that understands both text and voice. It reads messages, turns voice notes into text, and sends helpful replies fast. Great for support teams that want quick answers without switching tools.

When a chat arrives, the bot shows typing to set expectations. A switch checks if the message is text or voice. Voice notes are downloaded and sent to OpenAI to convert audio into text. Text and transcribed words are merged with labels like source type and message type so the reply can be tailored. An AI agent powered by GPT 4o uses a short chat memory of the last 10 messages per chat to stay on topic. The reply is cleaned to avoid Telegram formatting issues and then delivered. If someone sends an unsupported command, a friendly message explains what to do instead.

Setup is simple. You only need a Telegram bot and an OpenAI API key. Most fields are preconfigured, so you can go live quickly. Expect faster replies, less manual typing, and the ability to handle more chats with the same team size. Use it for common questions, order status checks, first line triage, or internal help for field teams.

• Listens to Telegram chat events and shows typing so users know a reply is coming.
• Detects if a message is text, voice, or unsupported and routes it correctly.
• Downloads voice notes and uses OpenAI to convert audio into text.
• Combines text and transcription with labels like source type and message type.
• Uses an AI agent with GPT 4o and a memory window of the last 10 messages per chat.
• Cleans special characters to prevent Telegram HTML parse issues.
• Sends a clear final reply and adds a thank you line that references the message type.
• Returns a friendly error when a command is not supported.

• Cut first response time to under 10 seconds
• Automate up to 70 percent of repeated questions
• Handle three times more chats with the same team
• Remove manual transcription for voice notes
• Avoid Telegram formatting errors by cleaning text

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Telegram and OpenAI. See the Tools Required section above for links to create accounts with these services.
3. In Telegram, talk to BotFather to create a bot and copy the bot token.
4. In the n8n credentials manager, create a Telegram credential using the bot token. Give it a clear name so you can find it later.
5. Open the Telegram Trigger node and select your Telegram credential. Leave updates set to all and save the node.
6. In your OpenAI account, create an API key from the API settings page.
7. In the n8n credentials manager, create an OpenAI credential with the API key. Select this credential in both the Chat Model node and the Audio Transcribe node.
8. Check the memory node settings. The session key should use the chat id and the window length should be 10.
9. Open the Set node and review the fields for combined message, source type, and message type. Adjust wording if you want different labels.
10. Turn on the workflow. In Telegram, start a chat with your bot and send the first message to activate it.
11. Send a text message to test. You should see typing, then a clean AI reply.
12. Send a voice note to test transcription. Confirm the reply reflects the transcribed text.
13. If replies are empty or slow, check your OpenAI usage limits and the node logs in n8n. Make sure the audio file id is present on incoming voice messages.
14. If nothing arrives, confirm the bot token is correct and that you started the chat with the bot in Telegram.