PIGONA
Translation for Discord

Live Voice Translation for Discord Voice Channels

Live Voice Translation turns Discord voice chat into translated text, optionally read aloud. How to start it, what it costs and what happens to your voice.

Live Voice Translation turns what people say in a Discord voice channel into translated text. Pigona joins the voice channel and posts each translation as a message in a text channel, next to the speaker's server name, with the transcript in the language that was spoken underneath. It can also read the translations out loud in the voice channel. Nobody is captioned until they choose it themselves, except people who can be heard through a captioned member's microphone.

Feature Info

How to start it

Join the voice channel first, then run /voice start in the text channel where the translations should appear.

Choose the language. Type to search, or leave it empty to pick from a list. The translations are posted in that language for the whole session.

Optionally choose the voice channel (default: the one you are in) and the channel for the translations (default: the one you ran the command in).

Read the private notice and choose: Start, caption me; Start, caption me and read aloud (only offered when the session reads aloud); or Start, don't caption me. You confirm that you are 18 or older, that you agree to the Terms and have read the Privacy Policy and, if you caption yourself, that nobody else can be heard through your microphone unless they agree. Choosing to caption yourself is your explicit consent to your speech being transcribed, translated and posted.

Pigona posts a public notice, in the translations channel and, where it can, in the voice channel's own chat, saying who started the session and whether they are captioned. Then it joins the voice channel.

The translations channel should only be readable by people who can see and join the voice channel, because otherwise what is said is published to people who were never in it. The voice channel's own chat always fits, and so does a private channel or private thread that only the people you invite (and moderators) can read. If more people can read it, Pigona warns you first and says who, and starts only if you confirm that these people may read what is said. Only a member with Manage Channels or Administrator in the translations channel can confirm this; a member who starts sessions through a role delegated in Pigona has to choose another channel. The public notice then says that people outside the voice channel can read the translations, and so does the private notice every member reads before they choose to be captioned. Pigona does not check that session again: your confirmation is about who can read the channel when it starts, and if the channel is opened up further later, the session goes on. A session started without that warning still ends if the channel is opened up later.

Before you start: you are responsible for telling the people in the voice channel that what captioned members say is transcribed, translated and posted as messages, and for the laws that apply where you are. Pigona's notice helps with this but does not replace it.

Checking and stopping

/voice status shows what is running: the language, the channels, how long it has been going, how many members are captioned, the Credits used and left, and whether reading aloud is on. /voice stop ends the session; the member who started it or anyone who is allowed to start sessions can run it. /voice help explains the basics in Discord.

A session also ends on its own at 2 hours, when everyone has left the voice channel for 2 minutes, when nobody who chose to be captioned has spoken for 10 minutes, when nobody has been captioned for 5 minutes (this includes a session in which nobody chose to be captioned), or when the server runs out of Credits. Pigona says why in the translations channel. A new session cannot be started within 60 seconds of the server's previous session starting or ending, or of the same member's previous start in any server.

Joining and leaving as a member

While you are in the voice channel, press Caption me on Pigona's notice, or run /voice caption-me. In every session, Pigona first shows you a private notice in your own Discord language (the 🌐 button on the public notice shows it in your language too) that explains what happens to your speech, and you choose Captions only, or Captions and read aloud (only in sessions that read aloud). Making that choice is your explicit consent to being captioned.

You choose in every session: Caption me always shows you the choice first. Pigona keeps a record of your latest choice only, as proof: a new choice replaces the earlier one (including its answer about reading aloud), and if you withdraw, the record shows when. It is kept for up to 12 months after your latest choice, and Pigona does not use it to caption you without showing you the choice.

Press Stop captioning me on the notice, or run /voice stop-me, at any time. It also works after leaving the voice channel. From then on your speech is not sent for translation, and captions of yours that are still waiting are dropped. Pigona's record of your latest choice then also shows when you withdrew it. Captions that were already posted stay in the channel until you or a moderator deletes them.

Deleting your captions: while a session runs and until 30 minutes after it ends, run /voice delete-my-captions, or press Delete my captions on the answer to Stop captioning me or on the notice Pigona posts when the session ends. Pigona edits each message to take out only your captions, or deletes it when nothing else is left in it, among the most recent 7,500 caption messages of a session and at most 100 messages per press (press again for the rest). It works after you stopped captioning and when you are no longer in the voice channel. It cannot change a message someone else has edited or one that is already gone, and it does not take back what others have already read, copied or been notified of, or speech already sent for translation. What other people said that was heard through your microphone was captioned as yours and goes with your captions. Credits already used are not refunded. After those 30 minutes, or once Pigona restarts, only a moderator can delete them.

Your consent lasts for the rest of the session, including about five minutes after you leave the voice channel, so a short disconnection does not need a new press.

Pigona does not separate voices: anyone who can be heard through a captioned member's microphone is captioned too, as part of that member's speech. If you caption yourself, use headphones and a private room, and mute your microphone or stop captioning if someone else becomes audible. Never press Caption me for someone else or pressure anyone to.

A member who has not pressed Caption me is not captioned. Discord sends Pigona the audio of everyone in the voice channel, but for them it is discarded without being decrypted: it is never transcribed, translated, sent on or stored. Pigona still sees who is in the channel and when someone is speaking, but not what is said.

Reading aloud

Pigona can also read the translations out loud in the voice channel, in a stock voice. It is on by default for every server, and it reads nobody aloud by itself: each member who presses Caption me chooses whether their own translations are read aloud. Members who chose captions only are never read aloud. Everyone in the voice channel hears the rest.

Turning it off. A member who is allowed to start sessions switches it off for the server with /voice settings speak enabled:false, and on again with /voice settings speak enabled:true. It is read when a session starts, so switching it on applies to sessions you start afterwards. Switching it off also stops it in a running session. The same switch is on the server's page in the dashboard, for members with the Manage Server permission or a Pigona Configuration or Admin role.

Those members can also switch it with an AI assistant they have connected to Pigona and allowed to make changes (see Connect Your AI Assistant).

What is read aloud. The translation of what members who chose it said, in the session's language. Pigona does not announce who is speaking, never copies or imitates anyone's voice, and does not read a caption that is already in the session's language. It needs the Speak permission in the voice channel and must not be muted there.

Languages and channels. Captions work in every language Pigona supports. Reading aloud is available in 46 of them; for any other language the captions work as usual and nothing is read aloud. Stage channels get captions but no reading aloud. /voice status shows when reading aloud is off and why.

When it switches itself off. If the server runs low on Credits, if reading aloud keeps failing, or if Pigona can no longer be heard in the channel, reading aloud switches off for the rest of the session and the captions go on.

Captions work for every language Pigona translates. Reading aloud is available when the translations are in one of the 46 languages below, in a computer-generated voice, and it costs the same Credits whatever the language. For any other language you get captions only. Chinese (Traditional) is read aloud in Mandarin and Portuguese in Brazilian Portuguese.

What it costs

Live Voice Translation uses the Translation Credits of the server it runs in, measured by the length of the speech in each translation that is posted. Captions cost 10 Credits per minute of speech. Reading aloud adds 40 Credits per minute of speech, charged only for lines that were actually read aloud: 50 Credits per minute of speech in all when every line is read aloud. Each translation counts for at least 4 seconds of speech.

A session needs at least 10 Credits to start (a minute of captions), or 13 when it reads aloud (a minute of captions plus one short line read aloud). A free server has 50 Credits a day, its whole shared pool. Reading aloud starts only when at least about 13 Credits are left that day (otherwise the session starts with captions only) and switches off a little before the Credits run out, keeping 5 Credits free for the captions plus the price of the next line, so on a free server it is at best a bit under a minute of speech a day, captions included. When the Credits run out altogether, the session ends.

Permissions

In the voice channel: View Channel and Connect to join. Speak and Use Voice Activity to be heard when reading aloud, and Pigona must not be muted.

In the translations channel: View Channel, Send Messages (Send Messages in Threads in a thread) and Embed Links. In a private thread, Pigona also has to be a member: mention Pigona itself in the thread once (the bot, not its role). If the reply to /voice start says Pigona cannot see an archived thread, reopen the thread and run /voice start again.

Optional: Set Voice Channel Status lets Pigona show a status line on the voice channel while a session runs, saying that Live Voice Translation is on. It is best effort: without the permission nothing is set and everything else works.

Channel permission rules override server-wide ones. A rule for @everyone that denies Connect, Speak or Send Messages in the channel also applies to Pigona unless a rule for Pigona's role allows it, so the server-wide permissions can look right while the channel still blocks it. If Pigona's role is missing permissions that Live Voice Translation needs, /voice start shows a button for someone with the Manage Server permission to restore them.

What happens to your voice

Pigona does not record or save voice. The audio of members who chose to be captioned is held in memory only while a passage is being translated, and is never written to disk. Silence is removed first, and only the speech is sent for translation to our AI translation provider, in the EU. If the provider's automated abuse checks flag a request, it may keep that speech for up to 90 days, and authorized staff may review it, solely to check for abuse.

Pigona stores no audio, transcript or translation, and creates no voice prints: it does not recognize anyone by their voice. When reading aloud is on, only the translated text of members who chose it is sent to turn it into speech, never their audio, name or User ID, and the generated audio is held in memory only until it has been played or dropped. The usage records Pigona keeps contain nothing that was said.

Pigona keeps a record of your latest Caption me choice so that it can show that you agreed: your Discord User ID, that choice and when you made it, and whether and when you withdrew it afterwards. A new choice replaces the earlier one, so earlier choices are not kept. It is kept up to 12 months after your latest choice, and deleted if you ask us to erase your data. It is never used to caption you without showing you the choice: you choose in every session.

Captions are ordinary Discord messages. Anyone who can read the channel can read them, and they stay there until someone with the right deletes them. Pigona keeps no copy of them. So that each member can delete their own, it keeps in memory only, while a session runs and for 30 minutes after it ends, a list of the caption messages it posted: message and channel IDs, and for each caption the speaker's User ID and its length, never the text or a name. The list is never written to disk and is lost when Pigona restarts; after that, removing a caption is up to your moderators, as with any other message. Stopping your captions or leaving does not delete captions that were already posted.

The full detail is in the Privacy Policy and the Terms.

Limits

Live Voice Translation is for members aged 18 or older. Each member confirms their age themselves; Pigona does not check it.

A session ends after at most 2 hours; start it again to continue. Only one session runs per server at a time.

It works in servers only, not in private calls.

A translation that is already too old by the time it is ready is skipped, so what you read stays close to what was just said.

Captions are machine output. They can be wrong, attributed to the wrong speaker, incomplete or contain words nobody said, for example on near-silence or background noise. Reading aloud can be wrong, mispronounced or played late. Do not rely on either in safety-critical, legal or medical situations, or as a record of what was said.

Troubleshooting

No captions, no sound, or a session that stopped? The troubleshooting answers in the FAQ cover each case. For permissions in depth see Common Error States, and for Credits see the Credit System.

This page is part of the Pigona Documentation, a complete guide to setting up and using Pigona - the context-aware AI translation bot for Discord. Pigona supports 100+ languages, Autonomous Translation, Reactive Translation, Relay Channels, and Smart Context Engine.

Related articles

← Back to all documentation · Return to pigona.app