Speaker attribution and turn-taking notes

Knowing what was said is only half the battle; knowing who committed to a deliverable is what makes meeting notes actionable. Oyma utilizes dual-channel audio separation and turn-taking heuristics to organize conversational flow.

The dual-channel separation advantage

Virtual conference calls feature a natural physical boundary:

• Channel 0 (Local Microphone): Your voice alone.
• Channel 1 (System Audio Loopback): Everyone else speaking on the call.

Because ScreenCaptureKit isolates these two streams into distinct audio buffers before Whisper inference, Oyma guarantees that your spoken words are never conflated with remote participant statements.

[Channel 0: Host Mic]    ──► [Speaker: You]
[Channel 1: SCK Loopback] ──► [Speaker: Remote Participants]

Renaming speakers in the transcript

When multiple remote participants speak on Channel 1, the engine separates turns based on acoustic pause boundaries. You can map these blocks to real team members in seconds:

1. Hover over a speaker header (e.g. Speaker 1).
2. Click the name label or press Return.
3. Select an attendee suggested from your Calendar invite (e.g. [[Sarah Connor]]), or type a custom name.
4. Check Apply to all occurrences in this transcript.

Action item ownership assignment

When AI summaries extract action items, speaker attribution ensures tasks are assigned accurately:

Extracted Action Item	Inferred Owner	Context Quote
Review database indexing RFC	[[Alex Chen]]	“I’ll take the lead on auditing the FTS5 query performance.”
Update customer security audit	[[Sarah Connor]]	“I can send the SOC2 packet by tomorrow afternoon.”

Related guides

• Pipeline mechanics in System audio pipeline and ScreenCaptureKit.
• Summary generation in AI meeting summary archetypes.
• Calendar matching in Calendar integration and attendee matching.