Glimpse v1.0.1

Improvements

• Improved UI stability.
• Improved crash reporting so startup and frontend issues are easier to diagnose.

Glimpse v1.0.0

Features

• You can now install Glimpse with Homebrew: brew install --cask glimpse-hq/tap/glimpse. This sets up the app and the glimpse command together. Updates still come through Glimpse itself.
• Added OpenRouter as a first class cloud speech provider, with its transcription models listed automatically.
• Library transcriptions can now detect speakers automatically when you use a provider that supports it (like Mistral). Turn on Detect speakers when importing or retranscribing and the transcript gets labeled by who's talking.

Changes

• Enabled 3rd party STT models now show in the models screen. This replaces the warning.

Improvements

• Serving models via API to other model providers is smoother.
• The API now automatically adds url prefixes if you forget.
• The CLI commands have been overhauled to be cleaner and show you all options.
• Many UI elements have been tweaked to be cleaner.
• Glimpse can now be auto refocused by searching it up in windows, instead of having to use the tray.
• (Nvidia) Long imported files now fill in their transcript and timestamps as they process, instead of only showing up once the whole file is done.

Fixes

• Fixed the CLI not being able to use different media formats.
• The models screen now shows which cloud provider and model is active, instead of leaving the name blank.
• Fixed remote speech model lists not loading from some self-hosted servers.
• Fixed Perplexity not working as a language model provider.
• The Custom speech provider no longer defaults to an OpenAI model that may not exist on your endpoint.
• The library will now correctly react to OS level pauses.
• On Windows, license activation links now work when Glimpse is already running, instead of being ignored.
• Fixed a startup crash for some modern CPUs on Windows.

Glimpse v0.9.9

Features

• Added Parakeet Unified (experimental): one English model that handles both fast file transcription and live streaming, so a single download covers what used to take two. It's about a quarter the size of Nemotron Streaming, faster, and adds word-level timestamps.

Improvements

• The Glimpse command-line tool can now do much more. Beyond transcribing files, you can browse and search your history, import and export library items, manage your dictionary and replacements, switch models, and check status, so you can script Glimpse or build your own integrations (Raycast, alfred, etc.). Run glimpse --help to explore, and add --json to any command for easy scripting.
• Tidied up model names so it's clearer what each one is.
• Older models can now be marked Legacy. They no longer show up to download, but if you already have one installed you can keep using it.
• You can now open the customer portal to manage your license straight from the account view, even when you're signed out.
• The start of your recordings no longer gets clipped: the mic now warms up as the pill appears, so your first word makes it in.
• Fixed a rare hang where transcription could stall for a while if the app you were in was slow to respond.
• Fixed library items that could come unstuck from their audio after an update.
• Licensing is more reliable behind the scenes.

Changes

• Nemotron Streaming is now Legacy, superseded by Parakeet Unified. It won't appear to download anymore, but if you already have it installed it keeps working.
• Updated onboarding for a more unified experience.
• Small UI changes.
• Small change to the UI icon for a cleaner look.

Fixes

• When using a streaming model, it could get mixed with ongoing library transcriptions if being done simultaneously.

Glimpse v0.9.8

Features

• Playing back transcripts in library view now show word-level timestamps. (You will need to retranscribe library items for this to take effect)

Improvements

• Local transcription is faster: Parakeet and Nemotron now use more performance cores, making them up to 50% quicker.
• Fixed a bug where live-streamed words could be blurred and improved streaming animations.
• The about page now has a new support page, you can export logs here too.
• Glimpse now uses Silero VAD for higher quality noise cancellation.
• You can now say Thank you.

Glimpse v0.9.7

Features

• Added support for using mouse buttons as keybinds.

Improvements

• The library import modal is now cleaner and supports more options.

Glimpse v0.9.6

Features

• Whisper models can now run on the Apple Neural Engine (ANE) on Apple Silicon Macs, making transcriptions faster and more power-efficient.
• Library overhaul: A new cleaner UI, support for assigning speakers, and better performance.

Improvements

• Much of Glimpse in this update has been optimized for an even smoother experience :)
• Local Whisper transcription is up to 30% faster after decoder tuning.
• Improved performance and reduced memory use during transcription, especially for longer recordings and cloud speech providers.
• Improved recording pill responsiveness and reduced unnecessary background work while recording.
• Improved History performance when loading recordings or importing transcripts.
• Parakeet now transcribes long recordings in chunks, improving transcriptions past around 5 minutes.

Changes

• Onboarding setup is down to one question and on Apple Silicon the Neural Engine encoder comes included automatically.
• Distil whisper models are now under experimental (They're not great at following dictionary rules).

Glimpse v0.9.5

Improvements

• The recording pill now moves out of the way when your cursor is near it.
• Models page revision: each model shows its quantization and installed versions are highlighted.
• What's New now renders full Markdown.

Changes

• Home greetings no longer change based on the day of the week.
• Recordings now stop on their own after 30 minutes.

Fixes

• The Models screen now updates right away when you install or delete a model, instead of only after reopening Settings.
• Cancelling a model download partway through no longer leaves an empty model folder behind.

Glimpse v0.9.4

Features

• Glimpse now supports all whisper models.
• Onboarding is now a quick guided setup: tell Glimpse which language you dictate in and whether to prioritize quality, balance, or size, and it picks and downloads the right local model for you.
• New model browser for choosing local models: search by name, filter by category, and see each model's size and supported languages at a glance. Quantization options (Q5, Q8, Full) are grouped under each model so switching between them is simpler.

Improvements

• Much of the UI has been cleaned up and improved to feel cleaner.
• Words spoken on your account card is now a running lifetime total, it counts everything you've dictated and is no longer based on the current amount of transcripts.
• Model downloads are now verified after they finish, so an interrupted or corrupted download won't leave you with a broken model.
• The transcription language list now clearly separates the languages your current model supports from the ones that need a different model.
• Auto Dictionary now works with online transcription providers, not just local models.

Fixes

• Fixed cancelling a dictation and immediately starting a new one occasionally pasting the cancelled text and glitching the recording pill.
• Fixed the Nemotron Streaming model downloading mismatched files, which left it broken and showing the wrong size.

Glimpse v0.9.3

Features

• Glimpse can now recover recordings if the app closes or crashes while you're still dictating, then save them to History when you reopen it.

Improvements

• Home now has a date header with rotating greetings and today's dictation stats.
• Setting shortcuts now works more like other apps: press the full combo, release all keys to save it, and the result no longer depends on which key you let go of first.
• Transcript history search is cleaner now, with quick sorting and time filters for finding older dictations faster.
• The license card in Settings → Account and onboarding now surfaces plan pricing more clearly, including per-seat Commercial licensing and volume discounts.
• Personalization mode handles long app and website lists better and looks a bit cleaner.

Changes

• Updated Glimpse Personal pricing: Personal is $24.99, and Commercial licensing starts at $19.99 per seat with volume discounts as your team grows.

Fixes

• Fixed cancelling a dictation while it's processing sometimes still pasting the text.

Glimpse v0.9.2

Features

• During onboarding, Glimpse can detect Aqua Voice, superwhisper, Wispr Flow, or Handy and import what you want, dictionary words, text replacements, personalization modes, shortcuts, language, launch-at-login, speech models, and past transcripts.

Changes

• Auto-pause media now works system-wide on macOS and Windows, not just with a handful of apps. In Settings → App you can pause playback completely, duck it to 10%, 25%, 50%, or 75%, or turn it off.
• Glimpse CLI can now be pointed at any whisper models, it is no longer locked to Glimpse specific ones.

Fixes

• Fixed some modifier-only shortcuts, like Fn or Option, occasionally not responding.
• Fixed HTML-style lists in transcript history showing as plain text.
• Fixed the Transcription Language dropdown in Settings opening behind the Edit Mode section.
• Cleaned up some UI elements.