Trace

Product Introduction

1. Definition: Trace is a native macOS menu-bar application that functions as a fully on-device AI meeting transcription and note-taking tool. It is a local speech-to-text engine and audio capture utility designed for privacy-conscious professionals.
2. Core Value Proposition: Trace exists to provide a fast, private, and frictionless way to generate searchable, markdown-formatted meeting transcripts without sending any audio data to the cloud. Its primary value is in delivering local transcription for Mac, offline meeting transcription, and privacy-first note-taking directly from your menu bar.

Main Features

1. On-Device Audio Capture & Transcription: Trace captures audio from your microphone and system audio as separate, discrete WAV files. It then processes this audio locally using a local speech model running entirely on Apple silicon (M-series chips). This means transcription occurs in seconds, not minutes, with no network latency or upload queues. The technology leverages Apple's Neural Engine for efficient, on-device AI transcription.
2. In-Flow Key Moment Flagging: During a recording, pressing the global shortcut ⌘K brings up a small input field. You can type a short note (e.g., "action item," "decision") and hit Enter. This key moment is timestamped and inserted inline within the final transcript. This feature enables live meeting annotation without switching contexts, creating a contextual transcript where important points are immediately visible.
3. Live Recap Pill & Menu Bar Management: Pressing ⌘? expands the recording pill into a live view showing the last two minutes of transcription interleaved with your flagged moments. All recordings are listed chronologically in the menu bar dropdown, allowing one-click access to copy, reveal in Finder, or re-open previous sessions. This creates a persistent meeting history that is always accessible but never intrusive.
4. Local File Export & Markdown Output: Every recording session is saved to ~/Library/Application Support/Trace/ as plain, open files: separate mic.wav and system.wav audio files, a clean transcript.md file, and a transcript.json for programmatic use. The markdown output is formatted with speaker labels, timestamps, and inline key moments, making it ready to paste into Notion, Obsidian, ChatGPT, or Claude for further processing or archiving.
5. Meeting-Aware Calendar Integration (Opt-In): An optional, read-only connection to Google Calendar allows Trace to use event titles for automatic transcript naming. It can also provide a quiet menu bar notification one minute before a meeting starts, prompting you to begin recording. This calendar-aware recording feature is off by default, respecting user privacy.

Problems Solved

1. Pain Point: The privacy and security risks of cloud-based transcription services (e.g., Otter.ai, Fireflies.ai) where sensitive meeting audio is uploaded and processed on external servers. Trace solves this with zero-data exfiltration and local-only transcription.
2. Pain Point: The disruptive workflow of manually taking notes or flagging important points during a fast-paced meeting, which takes focus away from the conversation. Trace solves this with its global shortcut for key moments (⌘K) and live recap view (⌘?).
3. Target Audience: Security-conscious professionals (lawyers, therapists, executives), remote developers and product managers in daily syncs, journalists and researchers conducting interviews, and any Mac user who participates in frequent video calls on Zoom, Microsoft Teams, or Google Meet and needs searchable records.
4. Use Cases: Client confidentiality meetings where NDAs prohibit cloud processing; daily engineering standups where action items need tracking; user interview synthesis where accurate, timestamped quotes are crucial; personal note-taking for lectures or podcasts played through system audio.

Unique Advantages

1. Differentiation: Unlike cloud-based competitors (Otter, Grain), Trace requires no bot to join the call, no account, and no subscription. Unlike other local recorders, it provides speaker-aware transcription and inline key moment flagging entirely on-device. Compared to simply recording, it delivers structured, searchable text, not just audio.
2. Key Innovation: The integration of a capable local speech recognition model that runs efficiently on Apple silicon, combined with a system-level audio capture that requires no virtual microphone or audio routing software (beyond macOS permissions). This creates a seamless, native macOS app experience for private transcription that is both powerful and simple.

Frequently Asked Questions (FAQ)

1. Does Trace work without an internet connection? Yes, Trace is designed for offline transcription for Mac. All audio capture, speech recognition, and transcript generation happen locally on your Mac. An internet connection is only required for the initial download from the Mac App Store and for the optional Google Calendar integration.
2. How does Trace capture audio from other apps like Zoom or Teams? Trace uses the standard macOS System Audio Recording permission. When you grant this permission, Trace can capture the audio output of any other application, such as Zoom, Google Meet, or Spotify, directly. It captures this as a separate track from your microphone input.
3. Where are my recordings and transcripts saved? All data is stored locally on your Mac in the folder ~/Library/Application Support/Trace/. Each session is saved as standard WAV audio files and Markdown/JSON text files. You own these files completely and can move, copy, version them in git, or delete them at any time.
4. Can I use Trace with Google Meet or Microsoft Teams without a bot? Absolutely. A core advantage of Trace is no-bot transcription. It captures your system's audio output, meaning it works with any video conferencing software, webinar, or even audio playing from a browser, without needing to add a third-party participant to your call.
5. What are the system requirements for Trace? Trace requires macOS 14 (Sonoma) or later and a Mac with Apple silicon (M1, M2, M3, or later). This is because the local speech model is optimized to run efficiently on Apple's Neural Engine, which is not present in Intel-based Macs.