Mic3 AI Notes: from recording to notes and tasks
A recording becomes more than a file to replay: the app prepares a transcript, notes and tasks. Audio is processed locally on the Mac.
App iconWho it is for
Leaders and teams working with long recordings and confidential audio.
What it enables
Processing produces a concise note, a full transcript and the original recording for review. Teams get usable materials without cloud audio uploads. AI output still needs human review; speed depends on hardware and local models.
Turn long wireless-microphone recordings into usable work output without Terminal, manual file assembly or third-party audio uploads.
- Transcription and summarization run on a specific Mac and depend on locally installed models.
- A mixed track cannot reliably identify speakers; transmitter recordings are better processed separately.
- The source WAV must never be modified or deleted, and repeated processing must be duplicate-safe.
- A native SwiftUI interface with drag-and-drop, queueing, progress and completed-note browsing.
- A local ffmpeg → Whisper large-v3-turbo → Qwen3 8B pipeline with dependency startup handling.
- A persistent library containing notes, transcript, SRT/VTT, JSON, source audio and processing logs.
04 System scope
What was designed and built
Native macOS application
Local AI audio pipeline
Notes and results library
05 Verified scale
A real 30-minute recording produced a 2,917-word transcript
The source is preserved with Markdown, TXT, SRT, VTT, JSON and logs
The arm64 app bundle passes local signature verification
Local-processing architecture, Python pipeline, SwiftUI app, background-process orchestration, file library, icon and build verification.