Send Audio 🔽

If you're generating audio in ComfyUI - TTS, music, sound effects - this is the node that gets it to your phone. It takes an AUDIO tensor and posts it to a Telegram chat as a voice note, a playable audio file, or a raw file. That "sent a voice message" mechanic is genuinely nice for TTS workflows: you end up with a chat history of every line the model read.

It's part of the ComfyUI Telegram Suite pack, and it's the sibling of Send Image(s) - same Telegram Bot wiring, same message/message_id outputs, one extra moving part: ffmpeg.

How it works

The AUDIO input follows ComfyUI's standard audio dict format (waveform + sample_rate, the kind of thing you get out of an audio-loading pack or a TTS node). The node renders it to WAV bytes with torchaudio, then branches on the send_as dropdown:

• Voice → converted to OGG (Opus) via ffmpeg and sent with sendVoice. This is the "voice message" style - the little bubble with the play button.
• Audio → converted to MP3 via ffmpeg and sent with sendAudio (the "music file" style with title/artist metadata).
• File → sent as-is with sendDocument (WAV, no conversion).

The quiet gotcha is in that middle column: converting to OGG or MP3 shells out to ffmpeg, which must be on your PATH. The README flags ffmpeg as a requirement up front, and it's the one dependency pip can't install for you. If you only ever use File mode you can skip it - but Voice and Audio will fail with a runtime error the moment they hit a machine without ffmpeg.

Inputs and outputs that matter

• bot, chat_id - from a Telegram Bot 🔽 node (your configured token + target chat).
• audio - the AUDIO tensor.
• send_as - Voice / Audio / File, described above.
• caption, parse_mode - text under the audio; parse_mode is None/HTML/Markdown/MarkdownV2.
• file_name (default "audio") - base name for the uploaded file.
• disable_notification (default True) - silent delivery, flip to False for an actual ping.
• protect_content - blocks forwarding/saving.
• message_thread_id - forum topic targeting; -1 means the main chat.

Outputs are message (raw API response DICT), message_id (INT - keep it if you want to edit this message later), and the trigger passthrough.

Installation

Same as every node in this pack - it all ships together:

cd ComfyUI/custom_nodes
git clone https://github.com/SwissCore92/comfyui-telegram-suite.git
cd comfyui-telegram-suite
pip install -r requirements.txt

or install "ComfyUI Telegram Suite" via ComfyUI Manager and restart. Then put your bot tokens and chat IDs in ComfyUI/user/default/telegram-suite/config.json and restart again. No model downloads; the pip deps are light (httpx, colorama, plus torchaudio). The one system-level requirement is ffmpeg.

Troubleshooting

• ffmpeg audio conversion failed - ffmpeg isn't on PATH. Install it system-wide (or drop an ffmpeg binary on PATH) and restart. Test with ffmpeg -version.
• "Telegram bot token ... was not found" - the config file was auto-created with placeholders on first run; fill in real values and restart.
• Voice/Audio send works, File doesn't, or vice versa - different code paths. File mode skips ffmpeg entirely; the other two require it.
• Your bot token lives in plaintext in config.json. Don't share that file.