| name | openai-tts |
| description | Generate short spoken audio files from text with OpenAI text-to-speech. Use when you want a Telegram voice-note reply or any local speech audio from text. |
| compatibility | Requires OPENAI_API_KEY or HERMIT_OPENAI_API_KEY, network access to api.openai.com, and a writable output path. |
OpenAI Text To Speech
Use this skill when you want to turn a short reply into a local audio file.
It is especially useful before calling send_telegram_voice_note.
What it does
- sends text to OpenAI text-to-speech
- defaults to a Telegram-friendly Opus file
- writes the audio file locally
- prints the output path
Usage
From the repo root:
node skills/openai-tts/synthesize.mjs --text "Short spoken update" --output workspace/inbox/telegram/reply.opus
Useful variants:
node skills/openai-tts/synthesize.mjs --text "I reviewed your plan" --output workspace/inbox/telegram/reply.opus --voice alloy
node skills/openai-tts/synthesize.mjs --text-file workspace/inbox/reply.txt --output workspace/inbox/telegram/reply.opus
node skills/openai-tts/synthesize.mjs --text "Quick update" --output workspace/inbox/telegram/reply.mp3 --format mp3
Defaults
- model:
gpt-4o-mini-tts
- voice:
alloy
- format:
opus
Recommended workflow for Telegram voice replies
- Draft a short spoken reply.
- Generate audio with this skill.
- Send it with
send_telegram_voice_note.
- Only add a caption if it helps.
Notes
- Keep spoken replies short and direct.
- If the user started with a Telegram voice note, voice is preferred for substantive replies.
- If text-to-speech fails, fall back to a short text reply and say so plainly.
