Text to speech

What it does

Liora can read typed messages aloud in a voice call, for people who can't or don't want to talk. It's off until a server's staff switch it on in her dashboard. Then she reads:

• a voice channel's own chat: what people in that call type in the call's chat (staff can switch this part off)
• a text channel the staff pick, if they pick one: what's posted there is read in the call its author is in
• /tts say: one line, in your call

Pick the voice she reads your words with using /tts voice. Anyone in her call can stop her with /tts stop.

Fair and safe

• She never listens. She joins calls deafened and only speaks. Nothing anyone says in voice reaches her.
• Her voice is made on her own computer, by a speech program called Piper. Your words aren't sent to any outside service to be spoken, and there's no AI involved.
• She won't read anything her Safeguard takes down (someone's personal details, IP-grabber links), anything deleted before or while she reads it (by you, the staff or her filters), messages from bots, or commands.
• Links, spoilers and code aren't read out: she says "a link", "a spoiler" or "some code" instead. Long messages are cut short after about 300 characters.
• One line at a time, a few at most. If someone types very fast, some of their messages are skipped, and if lots are waiting she asks people to give her a moment.
• Music comes first. With a song playing she speaks over it and the song dips while she does. She never moves away from music someone is playing.
• She leaves a call she joined only to read after five quiet minutes, or once everyone has left.

What she keeps

What she reads isn't kept: the words go to the speech program and the sound to the call, and both are gone once she has said them. She doesn't record the call or keep a list of what was said.