Voice & audio

The bot can speak, listen, and caption. All three features use OpenAI audio models and are metered separately from your monthly character cap. See Plans & billing for the allowances per plan.

Text-to-speech: /speak#

Turn text into a natural spoken audio clip, posted in the channel.

Options:

• text (required): what to say.
• voice (optional): Alloy, Echo, Fable, Onyx, Nova, or Shimmer. Your last choice is remembered, so you only set it when you want to switch.

The model automatically speaks whatever language the text is in, with a native accent. Metered in characters of input (roughly 1,000 characters per minute of speech): about 1 minute per month on Free, 10 on Pro, and 2 hours on Max. Your remaining allowance shows on /usage and the dashboard.

Speak in a voice channel: /voice (Max)#

Where /speak posts an audio clip, /voice joins your voice channel and speaks the text aloud into it. Be in a voice channel, then run /voice text:<what to say>; the bot hops in, says it, and leaves a few seconds after it's done.

Options:

• text (required): what to say out loud.
• to (optional): translate the text into this language first, then speak the translation, so you can say something in a language you don't speak, out loud.
• voice (optional): Alloy, Echo, Fable, Onyx, Nova, or Shimmer (your last choice is remembered).

Good to know:

• Max plan only: like /transcribe, it holds a live voice connection.
• It uses the same text-to-speech allowance as /speak (characters of speech); when a to language is set, the translation also counts against your character cap like any translation.
• If several people use it at once, lines queue and play in order. The bot leaves after about 10 seconds of silence.
• One voice channel per server at a time; it won't run while /transcribe captions are active (stop them first).
• The bot needs Connect and Speak in your voice channel.

Translate voice messages#

Right-click any voice message or audio clip and choose Apps → Translate. The bot transcribes the audio, shows you the transcript, and translates it into your language, all ephemerally (only you see it).

• Metered in seconds of audio: 1 minute per month on Free, 10 minutes on Pro, 2 hours on Max.
• Clips up to 5 minutes are supported.
• The translation itself counts against your character cap like any text translation (unless your engine is OpenAI, which is uncapped).

Live voice channel captions: /transcribe (Max)#

Join a voice channel and run /transcribe start. The bot joins, listens, and posts live captions into an auto-created thread: one message per spoken sentence, with the speaker's name, and a translation beneath it when a target language is set.

Options on start:

• language (optional): translate captions into this language. Defaults to your saved language, or your Discord client language.
• spoken (optional): the language being spoken. Setting this locks the transcriber to one language and makes captions far more accurate, so use it whenever everyone speaks the same language. Left unset, the bot learns each speaker's language as they talk, which suits multilingual calls.

Captions appear a moment after each person stops talking (about 1 to 3 seconds). Stop anytime with /transcribe stop (the starter, or anyone with Manage Server). Sessions also end on their own when everyone leaves the voice channel, or after 2 hours.

Good to know:

• One session per server at a time (a bot can only be in one voice channel per server).
• Transcription seconds and translation characters are billed to whoever started the session. Captions draw from the same monthly transcription allowance as voice messages, and the session ends on its own if it runs out.
• The server's access control (blocklist and required role) applies: blocked members are not captioned.
• Silence costs nothing; only actual speech is transcribed.

Allowances at a glance#

Allowances reset at the start of each calendar month, separately from your character cap. You get a heads-up at 80% and a notice when an allowance is used up.