Meeting Notes

Meeting Notes is a live audio recorder that transcribes what you say in real time. Start recording, and SupaNet uses OpenAI Whisper to convert speech to text as you speak. When you're done, save the transcript as a markdown artifact for future reference or sharing.

Why meeting notes are useful

Meeting Notes lets you capture conversations without manually typing:

• Live transcription — record audio and see words appear in real time; no silent pauses, no retype.
• Timestamps — each segment is timestamped so you can jump back to a specific part of the meeting.
• Live chat panel — ask the assistant questions about the meeting as it unfolds, or run a saved skill to summarize, extract action items, or anything else you need.
• Capture both sides of a call — optionally record the tab/app audio (Zoom, Meet, Slack huddle) alongside your microphone, so you capture what you said and what others said.
• Ambient ideas — optionally enable idea surfacing to get periodic suggestions of insights, questions, and action items while recording.
• Auto-generated summaries — when you save, SupaNet automatically distills a TLDR and action items from the transcript and any chat conversation, so your saved note is instantly organized and actionable.
• Transcript persistence — if you reload during a recording, your progress is automatically restored.
• Save as an artifact — once you've recorded and reviewed, save the transcript as a markdown artifact you can share, edit, or include in collections.
• No manual prep — just hit record and listen; transcription and summarization happen automatically.

Where to find Meeting Notes

On the sidebar under Assets, select Meeting Notes. You'll see a recording panel, a text input for the meeting title, and a live transcript view.

Setup requirement

Meeting Notes requires your workspace admin to configure the OpenAI API key. Go to Settings → Secrets and create a workspace secret named OPENAI_KEY with your OpenAI API key as the value. If transcription isn't configured, you'll see a warning banner on the Meeting Notes page before you record. Once the secret is added, transcription will work immediately.

Starting a recording

1. Enter a meeting title (e.g., "Weekly Standup" or "Client Call") — this is required before recording can start.
2. Optionally check "Also capture tab/app audio (Zoom, Meet, Slack huddle)" if you're on a video or audio call and want to record both sides of the conversation.
3. Click Start Recording. SupaNet requests permission to access your microphone, and if you enabled tab audio, you'll be asked to choose which tab or window to share (choose the one running Zoom, Meet, Slack, or the app you're calling from).
4. Once granted, the microphone icon pulses red and the status shows "Recording..." or "Recording (mic + shared audio)…".

The recording runs in the browser and never leaves your device until you save. Transcription happens every 15 seconds, so you see segments appear as they're processed.

Recording both sides of a call

When you check "Also capture tab/app audio":

• You'll pick the tab or window — when you click Start Recording, a system dialog appears asking which window or tab to share. Select the one running your meeting (Zoom, Meet, Slack, etc.).
• Always tick "Share tab audio" — the dialog offers options like "Share video" and "Share tab audio". Make sure "Share tab audio" is checked so SupaNet can hear what's playing.
• Graceful fallback — if no audio is shared from the tab, SupaNet automatically falls back to mic-only and lets you know.
• Both streams mixed together — your microphone and the shared audio are mixed and transcribed as one stream, so the transcript contains both what you said and what others said.
• Smart mic handling — when mixing mic and tab audio, SupaNet disables echo cancellation and noise suppression (which would otherwise duck your voice when the tab audio is loud) and boosts your microphone slightly so your voice stays clear and present while videos, music, or presentations play.
• Browser support — this feature works best in Chrome. Safari support is limited.

If you don't enable tab audio capture, SupaNet records your microphone only (the original behavior).

Transcript persistence

If you reload the page or navigate away during a recording, SupaNet automatically restores your in-progress meeting when you return — including the transcript captured so far and your meeting title and context prompt. The chat thread is also saved, so you can resume asking questions. Your progress is never lost.

Viewing the transcript

As you record, segments appear in the Transcript panel below the controls. Each segment shows:

• Timestamp — the time offset from the start of the recording (MM:SS format).
• Text — what was transcribed from that moment.

The transcript intelligently auto-scrolls as new segments arrive: when you're reading along near the bottom, it keeps you at the newest segment. If you scroll up to review earlier parts, the scroll position stays where you put it so you won't be interrupted. Once you scroll back toward the bottom, auto-scroll resumes.

Asking questions during the meeting

Click the Chat button to open a live chat panel beside the transcript. The assistant can see the current transcript and answer questions as the meeting unfolds:

• "What have we decided so far?"
• "List the action items"
• "Summarise the last few minutes"

The chat is saved as a real conversation, so it syncs across your devices and shows up in your chat list.

Ambient idea monitor

While recording, you can enable "Suggest ideas while recording" — a checkbox in the controls. Once turned on, SupaNet periodically checks the transcript and surfaces brief ideas, open questions, or action items that participants might miss:

• What it does: every minute, the assistant reads the new parts of the transcript and suggests one useful insight in a single sentence. If nothing stands out, nothing appears.
• Passive observation only: the monitor runs tool-free, so it only observes and never takes action.
• Opt-in: turn the checkbox on to start receiving ideas, turn it off to silence them.
• Live feed: ideas appear in a purple box below the controls as they arrive, building a list of insights you can review or discuss during the meeting.

This is useful for spotting discussion gaps or reminders that might otherwise get lost in a long meeting.

Using a context prompt

Before you start recording, you can optionally fill in "What should I do with this meeting?" — a set of instructions for what the assistant should do with the transcript when recording stops. For example:

• "Summarise the decisions and action items"
• "Create a to-do for each action item mentioned"
• "Extract the key technical decisions"

When you click Stop Recording, SupaNet automatically sends this prompt to the assistant with the final transcript. If you've selected a saved skill, it pre-fills the prompt with that skill's instructions.

This runs automatically after the final audio segment has transcribed, so you get results without an extra click. If there's no prompt or nothing was recorded, nothing runs.

Stopping and reviewing

Click Stop Recording to end the audio capture. SupaNet:

1. Stops the microphone stream.
2. Sends any remaining audio chunks for transcription.
3. If you set a context prompt, automatically sends it to the assistant.
4. Updates the status to "Recording stopped. Review and save."

Review the transcript in the panel. If you see errors or want to adjust the text (manual edits aren't supported yet — this is planned), you can clear it all and start over, or save it as-is.

Saving as an artifact

Once you've stopped recording and reviewed the transcript:

1. Click Save as Artifact. SupaNet formats the transcript as markdown with the following sections:

   • Meeting title with date — the title you gave it plus the recording date (DD-MM-YYYY format).
   • Ideas — any ideas surfaced during recording, listed as bullets. If none were captured, this shows "None captured."
   • Follow up — auto-generated action items and next steps extracted from the transcript and any chat conversation. If the generation fails or finds nothing, this shows "None."
   • TLDR — an auto-generated 2–4 sentence summary of what happened and what was decided. If the generation fails, this shows "Not generated."
   • Raw Transcript — the full transcript text with timestamps for each segment.
   • Chat (if applicable) — the full chat history if you had a conversation during the recording, so the saved artifact is a complete record.

   The TLDR and follow-ups are generated automatically from the transcript and chat by the assistant. This is optional — if generation fails or times out, the save completes anyway with empty sections so you don't lose your work.

2. The artifact is saved as a private markdown file (only you can see it initially).

3. SupaNet redirects you to the artifact editor, where you can:

   • Edit the transcript further.
   • Change the visibility to share it with your team.
   • File it into a collection.
   • Include it in a chat.

The artifact stores metadata recording the duration, idea count, whether it includes chat, and whether it has a generated summary, so you can filter or organize them later.

Viewing past meetings

Click the Past meetings tab at the top of the Meeting Notes page to see a list of all your saved meeting artifacts. The list shows:

• Meeting title — the name you gave it when recording.
• Date recorded — when the artifact was created.
• Duration — how long the recording was (e.g., "5m 30s").
• Chat indicator — a "chat" badge if the saved artifact includes conversation history.
• Ideas count — the number of ideas that were surfaced during recording (if any).

Click any meeting to open its full artifact, where you can read, edit, share, or file it into a collection.

Clearing a transcript

If you want to discard the current recording without saving, click Clear. SupaNet prompts for confirmation, then resets the transcript, title, and recording state so you can start fresh.

This action cannot be undone — any unsaved transcript is lost.

How transcription works

Transcription uses OpenAI Whisper, the same model that powers voice transcription across many apps. It supports:

• Multiple languages — Whisper detects the language automatically, but you can optionally specify one if transcription is going wrong.
• Noisy environments — Whisper is robust to background noise, but clarity improves the accuracy.
• Various audio formats — the browser captures webm/opus format, and Whisper accepts it.

Every 15 seconds of recording produces a transcription request, so even a long meeting is split into digestible chunks and processed efficiently. Each chunk is recorded as a self-contained audio segment and sent to Whisper independently, so you see the transcript build up in real time.

Microphone permissions

The first time you use Meeting Notes, your browser asks permission to access your microphone. Grant it to proceed. You can revoke permission in your browser settings at any time; the next recording will ask again.

Privacy note: Audio is sent to OpenAI's servers for transcription, but is not stored or used to train models (per OpenAI's policy). The transcript text is stored only in your SupaNet workspace as an artifact.

Using the transcript in chat

Once saved as an artifact, the transcript is a regular markdown file. You can:

• Share it — change visibility to workspace or public, or share a link.
• File it into a collection — organize with other meeting notes or project files.
• Chat with it — include the artifact's collection in a chat so the assistant can read and summarize the meeting, extract action items, or answer questions about what was discussed.

The assistant can also edit the transcript afterward to clean up errors, add headings, or format it further.