Trace: Offline Mac Meeting Transcripts with Real-Time Flagging

Trace is a local-first transcription tool for macOS that captures microphone and system audio to generate Markdown transcripts entirely on-device. Unlike traditional transcription services, Trace operates without accounts, cloud uploads, or meeting bots, ensuring that audio and text data never leave the user's Mac.

On-Device Transcription and Privacy

Trace utilizes a local speech model running on Apple silicon to transcribe meetings in seconds. By eliminating the need for servers, API keys, or telemetry, the application ensures total data privacy.

Key technical and privacy specifications include:

  • Hardware Requirements: Requires macOS 14 or later and Apple silicon.
  • Data Storage: Every session is saved in the Finder as plain audio (.wav) and Markdown (.md) or JSON files, allowing users to version them in git or move them into personal knowledge bases like Obsidian or Notion.
  • Permissions: The app requires System Audio Recording permission to capture audio from other applications but does not access the user's screen.

Real-Time Annotation and Workflow

Trace distinguishes itself from passive transcription tools by introducing an active annotation layer that allows users to mark important moments during a live call.

Key Moment Flagging

Using the shortcut ℝ K, users can type a short note mid-call. This note is inserted into the final transcript at the exact timestamp of the event. This allows users to create a structured record of a meeting as it happens, which simplifies later review or AI-assisted summarization.

Live Recap

The ℝ ? shortcut triggers a "pill" interface that expands to show a live transcript of the last two minutes of the conversation, interleaved with any flagged key moments. This provides a real-time reference for users who may have missed a detail without breaking the flow of the meeting.

Integration and Usability

Trace is designed to be keyboard-driven and unobtrusive, functioning as an overlay that can be summoned over any application, including Zoom, Microsoft Teams, Google Meet, and Slack.

  • Calendar Integration: Users can optionally connect their Mac calendar or Google Calendar (read-only) to automatically name recordings based on event titles. The app provides a reminder one minute before a meeting starts.
  • Speaker Identification: The tool includes speaker-aware capabilities to identify different participants and assign them unique identifiers in the transcript.
  • Keyboard Shortcuts: Core actions are mapped to shortcuts (ℝ R for record, ℝ K for key moment, ℝ ? for recap, ℝ P for pause, and ℝ H to hide), all of which are remappable.

Community Insights and Technical Discussion

Users on Hacker News highlighted several points regarding the utility and limitations of local transcription tools:

Value Proposition

Many users praised the one-time fee model over the subscription-based SaaS model. One user noted that the rise of distilled on-prem models is making it more tenable to build complete, standalone software without relying on cloud infrastructure.

Technical Challenges

Community members pointed out common pitfalls for offline transcription apps, including:

  • Crash Recovery: The risk of losing recorded audio if an app crashes before saving.
  • Disk Space: The accumulation of large .wav files over time.
  • Microphone Bleed: The issue of duplicate transcriptions when system audio is picked up by the microphone.

Feature Requests and Concerns

Users expressed interest in further enhancements, such as:

  • Language Support: Questions regarding the accuracy of the model for languages other than English (specifically German).
  • Legal Compliance: Concerns regarding the legality of recording in two-party consent states/countries.
  • Integration: Requests for Microsoft Teams/Outlook calendar integration and the ability to label speaker voices across different meetings.

Sources