Setup

Send Voice Messages in Telegram with ElevenLabs

A quick, deterministic setup to get your agent sending voice replies instead of text-only messages.

If your agent says “I need a TTS tool” or keeps replying in text even after you added ElevenLabs, this guide fixes the exact setup loop.

1) Confirm your base setup first

Most common miss: users copy a library voice URL/ID without adding it to Your Voices first. That usually fails in runtime.

2) Ask the agent to wire voice output explicitly

Set up Telegram voice replies using ElevenLabs. Do this now: 1) Verify ElevenLabs key is present and valid. 2) Verify voice ID is accessible to this key. 3) Enable voice output for Telegram replies. 4) Run a tiny smoke test and send one short voice message: "Voice test from your agent." Return: - provider used - voice ID used - any exact error text if failed.

3) Run a two-turn smoke test

  1. Send: "Reply to this with voice only: hello from Telegram test."
  2. Then send a second prompt with different wording.

You want two successful voice responses in a row. One success can be a false positive from stale state.

4) If it still sends text only

Copy/paste recovery prompt

Voice recovery mode. Goal: Telegram must send voice responses reliably. Rules: - Keep responses short. - If failure occurs, return exact provider error text. - Do not claim success without sending an actual voice reply. Steps: 1) Validate ElevenLabs key. 2) Validate voice ID access. 3) Send one 5-10 second voice reply in this chat. 4) Send a second voice reply. 5) Return provider + voice ID + proof of both sends.

Bottom line: treat this as a key+voice-access verification problem, not a prompting problem. Once key and voice ID are truly valid, Telegram voice replies become reliable.