Meeting Lifecycle

Every Chronote recording follows a predictable lifecycle. Understanding these stages helps you get better notes and troubleshoot issues when they arise.

Stage 1: Start

A meeting starts in three ways:

• Manual: A user runs /startmeeting while in a voice channel, or right-clicks themselves, Chronote, or someone in their voice channel and selects Apps -> Start meeting. The slash command can optionally include a context description and tags.
• Auto-record: A user joins a voice channel that has auto-recording enabled. Chronote starts recording automatically and posts an "Auto-Recording Started" embed.
• TTS-only: A user runs /say, or joins a channel with automatic chat-to-speech enabled. Chronote joins only to play TTS and does not create a recorded meeting.
• Personal upload: A user uploads an existing audio or video file from the web portal. Chronote creates a personal meeting owned by that user.

At this point, Chronote joins the voice channel and begins capturing audio from each participant separately. A recording message appears in the resolved notes channel with controls. For manual starts, the destination order is an optional notes-channel choice, the voice-channel override, the server default, then the invocation text channel only when nothing is configured. Context-menu starts use the saved defaults too. Auto-recording requires a configured destination and has no invocation-channel fallback.

The /startmeeting response is private, even when the recording is posted in the same channel. After Chronote publishes the recording message, View recording opens that exact message. The one-meeting choice accepts ordinary server text channels and does not change saved settings or grant channel access.

An unavailable or inaccessible configured destination blocks the start without fallback. Chronote fixes the destination when recording starts: its controls, processing updates and final notes stay there even if an administrator changes settings during capture.

Constraints: Only one meeting can be active per server at a time. The user who starts the meeting must be in a voice channel. The server must have available meeting minutes on its plan.

Stage 2: Capture

While a recorded meeting is active, Chronote captures three streams of data:

• Voice audio from each participant as individual audio snippets. Audio is processed through a noise gate that filters silence and very quiet segments.
• Chat messages from the text channel, including any images shared (which are later captioned by AI).
• Attendance tracking who joins and leaves the voice channel.

During capture, the meeting embed stays pinned with buttons for End Meeting, Edit Tags, and a link to the Live Transcript on the web portal (if configured).

TTS-only sessions skip capture entirely. Chronote does not subscribe to user voice, record bot audio, transcribe speech, store chat logs, generate notes, or upload meeting artifacts. The bot leaves when the voice channel empties, after a period without spoken TTS activity, or when an authorized user runs /tts stop or /leave.

Stage 3: Process

Processing begins when the meeting ends. The embed updates to show "processing" status. This stage involves several steps:

1. Transcription: Each audio snippet is sent to OpenAI's transcription model with a prompt that includes your dictionary terms and context. Quality checks filter out noise, prompt echoes, and low-confidence segments.
2. Finalized audio verification: After mixed audio is built, Chronote can run a final verification pass against the full recording and apply only high-confidence hallucination fixes before notes generation.
3. Transcript cleanup: A language model pass cleans up the raw transcription for readability.
4. Notes generation: The full transcript, chat log, participant roster, server/channel/meeting context, dictionary entries (with definitions), and recent meeting history are sent to a language model that produces structured meeting notes.
5. Artifact upload: The audio recording, transcript, and chat log are uploaded to cloud storage.

The processing step typically takes 1-3 minutes depending on meeting length.

Stage 4: Publish

Once processing ends, Chronote posts the available results to Discord:

• A Meeting Summary embed with: meeting name, one-sentence summary, start/end times, duration, attendees, voice channel, and tags.
• One or more Meeting Notes embeds containing the full generated notes (paginated if long).
• Action buttons: Open in Chronote (web portal link), Helpful / Needs work (feedback), Suggest correction, Rename meeting, Edit Tags.

The meeting is also saved to your meeting history in the database, accessible through the web portal and the /ask command.

A completed meeting means that processing ended. It does not guarantee that Chronote generated notes or produced a complete transcript. If a Discord or personal recording contains no usable speech, Chronote finishes without generating notes and explains why. If some audio could not be transcribed but usable text remains, Chronote generates notes from that text and warns that the notes may be incomplete. A processing failure remains separate from both outcomes.

If Notion automation is enabled, Chronote exports completed meeting notes after they are saved. Server automation uses the server manager's configured destination. Personal automation uses the personal meeting owner's Notion destination.

Interrupted recordings

If recording is interrupted, Chronote can recover verified saved audio or text without joining the voice channel again. Recovery follows the meeting's original transcription and notes settings. It does not generate notes when notes were disabled, and it does not invent notes when no usable saved speech remains.

Recovered partial notes keep an Interrupted status. A warning explains that they cover only the saved recording. The warning remains after notes edits and corrections, and appears in Discord, the library, shared meetings, copied or downloaded notes, Remote MCP, and Notion exports. Known transcription gaps still carry a warning even when recording itself finished.

The library shows recovery progress or when processing or delivery needs review. If saved text and audio cannot be combined safely, Chronote retains the content, marks recovery as needing review, and pauses automatic notes delivery and export. Confirmed saved audio duration describes preserved audio, which may contain gaps. It does not prove the full meeting was recorded. If the recording's exact end time is unknown, Chronote leaves it unknown.

If the live page cannot confirm the recording status, it shows Status unavailable instead of Complete. Use Retry to check again, or open the meeting library for saved content and recovery status.

Stage 5: Revise

After publishing, authorized users can refine the notes:

1. Click Suggest correction on the meeting summary.
2. Describe what should change in the modal (up to 1500 characters).
3. Chronote generates a minimal correction using the transcript as ground truth and shows a line diff.
4. An authorized user (meeting creator, or anyone with Manage Channels for auto-recorded meetings) reviews and clicks Accept & update or Reject.

Accepted corrections update the notes in place, increment the version number, and record the edit in the suggestion history. Each correction builds on the previous version, so the AI has full context of past edits.

Auto-record cancellation

If an auto-recorded meeting produces too little content (e.g., someone briefly joins then leaves), Chronote cancels the meeting instead of generating notes. The start embed is replaced with an "Auto-Recording Cancelled" message showing the reason.

Meeting duration limit

Meetings are capped at 2 hours. If the limit is reached, the meeting ends automatically and processing begins.

Weekly minutes limit

Each subscription tier includes a weekly meeting minutes allowance. If the limit is reached, new meetings cannot start until the next billing cycle. A warning is posted when the server approaches or hits the limit.

Context and quality factors

Several factors influence the quality of transcription and notes:

Factor	How to configure	Effect
Server context	/context set-server	Helps the AI understand your team, project, and domain
Channel context	/context set-channel	Adds channel-specific context (e.g., "design reviews")
Meeting context	/startmeeting context: option	Describes this specific meeting's topic
Dictionary terms	/dictionary add	Corrects spelling of names, acronyms, jargon
Audio quality	Participant microphone settings	Clearer audio produces more accurate transcription
Speaking volume	Participant microphone levels	Very quiet audio may be filtered by the noise gate