Voice meeting recordings

Description

User Guide

Record a voice meeting and automatically get the audio, a timestamped transcript, a summary, the list of decisions and everyone's speaking time. Designed for moderation teams, weekly check-ins and project meetings.

What it does

• Records a voice channel speaker by speaker, each voice on its own track.
• Produces a timestamped transcript, with each line labelled with its author's name.
• Writes a summary and a list of action items and decisions, in the server's language.
• Computes each participant's speaking time.
• Posts the minutes in the channel where the command was run, with the transcript attached.
• Keeps the audio and the minutes in the dashboard, with a built-in player.
• Deletes recordings automatically once the retention period has passed.

Quick start (Admin)

1. First of all: tell the participants the meeting is going to be recorded, and wait for their agreement.
2. Join the voice channel in question.
3. Run /record start from the text channel where you want to receive the minutes. The bot joins the voice channel and confirms with a "Recording in progress" message.
4. Hold your meeting as usual.
5. Run /record stop. The bot leaves the voice channel and announces how many speakers were captured.
6. Wait: the transcript and the minutes are computed in the background, then posted in the channel.
7. Find the recording at any time in the Recordings module of the dashboard.

Commands

Command	Description
/record start	Starts recording the voice channel you are in.
/record stop	Stops the recording, transcribes it and generates the minutes.

Settings from the dashboard

The Recordings module contains no settings: it is a history view, reserved for the server's administrators.

• The list of meetings: date, person who started the recording, voice channel, duration, and whether an audio file is available.
• A meeting's page: duration, number of speakers, number of speech segments, audio player, summary, action items and decisions, speaking time per participant, and the full formatted transcript.
• A Delete button that permanently erases the meeting: the page, the transcript and the audio file.

Good to know

Recording people is not a trivial matter

Always tell the participants before running /record start, and make sure they agree. The bot does not ask anyone for consent on your behalf:

• it sends no warning to the members already in the voice channel;
• it does not warn people who join the voice channel after the start — they will be recorded without ever seeing the confirmation message, which is posted in the text channel;
• no particular permission is required to run /record start: by default, any member can do it. If that is not what you want, restrict the record command to your trusted roles from Server Settings → Integrations.

Depending on your country and the nature of the meeting, recording a conversation without telling the participants may be illegal. Announce it out loud at the start of the meeting, and leave people the option to decline.

File retention

What the bot keeps, and for how long:

• The mixed audio file, the full transcript, the summary, the action items and the speaking time are saved in the database so they can be viewed from the dashboard.
• Recordings are purged automatically after 30 days by default (a daily check deletes the record and its audio file). This period is a setting owned by whoever hosts the bot: if you self-host it you can adjust it; on the TennoCore instance the default value applies.
• The individual per-speaker audio tracks are temporary: they are deleted as soon as processing ends. Only the overall mix is kept.
• You can delete a meeting at any time from the dashboard, without waiting for the purge: deletion also removes the audio file from disk.
• The transcript is also posted in the Discord channel as an attached file. That message, the bot will never delete: if you want it gone, erase it yourself, or start your recordings from a staff-only channel.

Other limits worth knowing:

• Only one voice use at a time per server. If the bot is already in a voice channel (music, or another recording), /record start refuses to start. Stop the music first.
• You must be in a voice channel to start the recording: the bot joins yours.
• The minutes arrive later. The transcript is computed on the server's CPU, which takes time — expect several minutes for a long meeting. The completion message arrives on its own, no need to run the command again.
• A meeting that is too short gets no summary. Below about thirty transcribed words, the bot deliberately gives up on summarising rather than inventing content; the transcript is still provided.
• The summary may be unavailable if the AI service that writes it cannot be reached. The audio, the transcript and the speaking time are still produced normally.
• If nobody speaks, there is nothing to transcribe: the bot tells you so and creates no record.
• The summary and the action items are written in the language configured for the server.
• The transcript is provided as an attachment within Discord's file size limit; for a very long meeting, it always remains fully viewable in the dashboard.
• Speech recognition is not perfect: proper nouns, jargon and overlapping voices produce errors. Proofread the minutes before using them as a reference.