Hermes Agent emits an ordinary audio file in Telegram, not a circular voice bubble, and many times it's not that the Telegram API is invalid, but that you are using the default Edge TTS without installing ffmpeg. The official Telegram documentation makes it clear: Edge TTS produces MP3 by default, which is converted to Telegram's more friendly Opus voice format by using ffmpeg.
How to judge if this is the problem
- If you are using Edge TTS and the message can be sent, but it is only displayed as a rectangular audio bar, 80% of the conversion is not done.
- If you switch to a more natively suitable output like OpenAI or ElevenLabs, this step is usually not necessary.
How to repair
To install ffmpeg on the machine:
- Ubuntu / Debian：
sudo apt install ffmpeg - macOS：
brew install ffmpeg
After installation, the voice reply is retriggered, and Hermes Agent can convert the Edge TTS results into a format that is more like a native voice message.
So let's not rebuild the bot as soon as this question comes up. As long as the text is normal and the voice can be sent, but the style is wrong, then it is enough to add ffmpeg first.
Official open source address: https://github.com/NousResearch/hermes-agent; Official document entry: https://hermes-agent.nousresearch.com/.
