Skip to content
TakoAPI
Back to skills

percept-speaker-id

by jarvis563

Coding Agents & IDEs
284Downloads reported by clawskills.sh
0Views
0 likesSign in to like this skill
—Not yet rated
Your rating

Brief

Speaker identification and management for multi-person conversations.

Description

What This Skill Does

Tracks speaker identity in multi-person conversations by mapping anonymous speaker labels to real names. Controls which speakers can trigger voice commands based on owner and approval status. Integrates with Omi pendant hardware to identify the primary wearer.

Combines hardware-based primary speaker detection with a configurable registry, avoiding the need for a fully automated voice recognition system while still supporting per-speaker authorization.

When to Use It

  • Attributing quotes to specific people in meeting transcripts
  • Restricting voice command access to approved household members
  • Reviewing who spoke when in a recorded multi-person session
  • Setting up a new speaker profile after adding a household member
  • Auditing which unknown speakers were present in a conversation

Example Workflow

INPUT

User asks: who said the last command?

AGENT
  1. 1

    Receive transcript segment with SPEAKER_01 label from Omi

  2. 2

    Look up SPEAKER_01 in percept/data/speakers.json

  3. 3

    Resolve label to name 'Rob' with approved status

  4. 4

    Check authorization level to confirm command was valid

  5. 5

    Return speaker-attributed transcript entry

OUTPUT

'Rob' triggered the command at [timestamp], authorized as an approved speaker

Requirements

  • Omi pendant device
  • percept-listen skill installed and running

Installation

clawhub install jarvis563/percept-speaker-id

Run the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.

Links