Changelog

What's new

Every version of Safeclip, documented.

Latest

Easier Mac setup

v1.0.1

If tool setup failed with a message about cmake or Whisper, this update fixes that. Safeclip now includes the transcription engine in the app, so you do not need developer tools, Terminal, or Homebrew.

What's new

  • macOS (Apple Silicon & Intel): Transcription engine is bundled with the app — no compile step during onboarding
  • Clearer setup screen: “Transcription engine” and “Language model” instead of two confusing “Whisper” rows
  • More reliable installs: Setup works when the app is opened from the Dock or Applications folder, not only from Terminal

What's unchanged

  • The language model (~142 MB) is still downloaded once on first use (same as before)
  • Clips, highlights, FFmpeg, yt-dlp, and Ollama behavior are unchanged

Who should update

Anyone on 1.0.0 who saw errors installing Whisper or cmake during Set up your tools, or anyone who wants the simplest Mac install experience.

Introducing Safeclip

v1.0.0

This is the inaugural public release of Safeclip, a private and local-first clipping and transcription application built to run fully offline. Built for sensitive work, Safeclip is designed for professionals who work with confidential recordings and cannot use general-purpose cloud tools.

Account & licensing

  • Free tier: sign in and use the full desktop app at no cost
  • Optional $99 lifetime license; plan status in Settings → Profile
  • 14-day offline grace after successful verification
  • Web signup logs in immediately for desktop users

Auto-updates

  • Update checker on launch and every 6 hours
  • Signed updater artifacts for macOS, Windows, and Linux

Processing

  • Highlight truncation fix for long-form Deep analysis (57+ min podcasts)
  • Clip depth selector (Standard / Deep) with reanalyze support
  • Speaker labels v2 (pause-based heuristics; imperfect on panels/overlapping speech)

Desktop UX

  • Branded macOS DMG installer
  • Project search, library recovery, drag-and-drop import
  • Export folder onboarding step
  • Custom window chrome and sidebar redesign

Requirements

  • Safeclip account (register free)
  • Local Ollama + model for highlight detection
  • Whisper model downloaded on first use (~140 MB)

Known limitations

  • Speaker labels are heuristic, not broadcast-grade diarization
  • Windows installers are not EV-signed (SmartScreen may warn)

Design Partner Release

v0.2.1

Safeclip is now in the hands of our first design partners. This release focuses on onboarding reliability, smarter local AI, and the foundation for account management.

  • Design partner invite and registration flow with email delivery
  • Ranked local model selection — Safeclip now picks the best available Ollama model for highlight extraction, preferring Llama 3.x and deprioritizing coding-specific models
  • Chunked highlight analysis for recordings over 12 minutes
  • Account sync, download history, and in-app feedback
  • Export platform preferences and audiogram style settings
  • Admin dashboard with user editing, access request management, and release notes CMS
  • Download page now auto-detects your platform and links directly to the latest build
  • Fixed design partners being unable to complete registration
  • Fixed admin Users tab only showing admins

Audiograms, Highlights, and Local AI

v0.2.0

A major feature release bringing AI highlight detection, audiogram export, a tiered scoring system, and the LLM provider abstraction that makes Safeclip model-agnostic.

  • AI-powered highlight detection with configurable clip length and count
  • Tiered highlight scoring with depth selector
  • Full audiogram export — cover art, speaker photo, word-highlight captions
  • LLM provider adapter layer — swap models without changing your workflow
  • Drag-and-drop file import via native Tauri
  • QuickTime-style timeline lock during clip drag
  • Thumbnail generation on export
  • Onboarding flow prefers language models over embedding models in Ollama setup

Private Alpha

v0.1.0

The first working build of Safeclip. Local Whisper transcription, manual clip selection, and MP4 export — entirely on your machine.

  • Local Whisper transcription — audio never leaves your machine
  • File import: MP4, MOV, MP3, WAV
  • Manual clip selection from transcript
  • MP4 export
  • macOS Apple Silicon and Intel support

Ready to try the latest build?