Back to selected work
Local-first audio productWorking prototype Evidence reviewed 2026-07-16

EchoPanel

A macOS recording and transcript workflow designed around local capture, private processing, and finding the exact moment later.

Primary user

People recording meetings, interviews, or research sessions who want searchable memory without defaulting to cloud storage.

My role

Product design and prototype implementation across capture flow, local transcription, transcript navigation, and retrieval.

Current outcome

Built a working native-product direction that connects capture, transcription, and retrieval; system-audio setup and production packaging remain active work.

Visual evidence

EchoPanel local audio workflow from native capture through local transcription, searchable timeline, and timestamp retrievalWorkflow map
Workflow map of the prototype direction. It separates the working capture and retrieval concept from system-audio setup and production packaging still in progress.

What exists now

Current implementation boundary

This list is intentionally narrower than a product roadmap. It describes the current working surface used for this case study.

Native macOS product surface

Recording workflow

Local transcription experiments

Transcript navigation and search direction

Export-oriented local storage design

Key product decisions

Judgment is more useful than a technology list.

01

Make capture a lightweight native action.

The product only becomes useful when recording starts with very little friction.

Trade-off

Native audio permissions and system-audio capture require platform-specific handling.

02

Prefer local storage and inference where practical.

Meeting audio can be sensitive, and local-first behaviour creates a clearer trust boundary.

Trade-off

Model size, performance, and hardware variance become product constraints.

Constraints

  • macOS microphone and system-audio permissions
  • Transcription latency on local hardware
  • Clear failure states when capture sources are unavailable

Technologies used

SwiftAVFoundationCoreAudioWhisperSQLite