Changelog

No notable changes in this release.

Improvements

• The Debug section in Settings now shows whether Google Analytics is enabled.
• Google Analytics now records app and overlay launches, page views, recording and transcription outcomes, and model download starts and results.

New Features

• Live audio-level indicators now appear beside available microphones in Settings, making it easier to check which input is picking up sound.
• A “Top pick” label highlights a recommended speech recognition model to help you choose.
• Usage analytics now records app screens and feature interactions to help the team understand how the app is used.

Improvements

• Audio visualizers respond more accurately to sound, with smoother and clearer displays.
• Microphone lists now animate when shown or hidden, and disconnected inputs are easier to identify.
• Model management is clearer: installed models have a more visible status and delete control, and the faster Whisper model has a clearer name.
• Navigation icons and spacing have been refined, and the OiPer option now uses updated premium branding.
• Settings use shorter, clearer labels for model preloading and history retention.

Improvements

• Audio recording now supports a wider range of microphone formats, helping it work more reliably across devices.
• Recording stops as soon as you end a session, while audio processing continues in the background.
• Audio device selection and playback are more reliable.

New Features

• Customize the recording overlay for each workflow with a wave, bars, block, or line style. You can also choose to show a workflow’s name.
• History details now compare speaking time with estimated typing time and show how much time you saved.
• In debug builds, History details includes an Advanced view with recording information and a step-by-step timing breakdown.

Improvements

• Choose whether local transcription models load when the app starts or when recording begins. Models loaded on demand are released when transcription finishes, helping save memory.
• On first launch, the app enables launch-on-login by default. You can change this in Settings.
• Settings are reorganized, with helpful explanations for advanced options and sound-effect and history-retention controls grouped with related settings.
• Workflow navigation and profile controls are clearer, including when no additional workflows are configured.

New Look and Feel

• Fresh visual refresh across the whole app: refined light and dark color palettes with better contrast and readability, softer borders, and rounder, smoother corner shapes on cards, buttons, and menus.
• The sidebar, tabs, tooltips, and buttons all have updated styling, with smooth animated highlights that follow your selection.
• Danger-colored actions (like deleting) are now more visible.

Workflows (formerly Profiles)

• Profiles are now called Workflows everywhere. The sidebar has a dedicated Workflows entry, and each of your app-specific workflows sits neatly beneath it in a tree layout.
• The old profiles list page is gone — clicking Workflows takes you straight to your main workflow, which now lists your other workflows below it along with the apps and shortcuts assigned to each. You can also create a new workflow right from there.
• When assigning apps to a workflow, the picker now finds every app installed on your computer — including per-user installs on Windows and Flatpak/Snap apps on Linux — with proper app icons.
• Duplicating a workflow now starts clean: the copy comes without shortcuts or app assignments, so you can set them fresh.
• Fixed: the Workflows item in the sidebar stays highlighted while you're browsing any workflow.

Redesigned Dictation Overlay

• The overlay that appears while dictating has a completely new look: a compact pill with a microphone icon, a live audio wave drawn as glowing dots, and your recording timer.
• It now "powers on" with a quick reveal animation when it appears.
• While your speech is being transcribed or enhanced, the wave shows a scanning sweep and gentle twinkles instead of a spinner.
• New: if you hold the shortcut but stop speaking, the wave gradually shifts color from cool blue to amber to red — so you always know OiPer is still listening.

Overview Dashboard

• New at-a-glance stats: words dictated, estimated time saved compared to typing, your speaking speed, and total dictations (including how many failed).
• Time ranges are now simple tabs (7D / 30D / 90D / All) at the top of the page.
• The "Tip of the day" banner is replaced by a shortcut banner showing your hold-to-dictate and press-to-toggle keys in full, readable form — with a "Change shortcuts" button.

New Features

• ElevenLabs now works as a cloud voice transcription provider. You can route your voice dictation through ElevenLabs for speech-to-text, alongside the existing cloud providers.
  • Note: ElevenLabs does not offer a "Translate to English" mode, so that language option can't be used with it — the app will tell you if you try.

Bug Fixes

• Hotkeys now behave more reliably while a recording is in progress on Windows and Linux. Previously, keys held to trigger or control a recording could be unexpectedly cleared mid-recording if a key release wasn't detected properly — this safeguard was already in place on macOS and now applies to all platforms, so hotkey behavior is consistent everywhere.

Improvements

• Grammar corrections now apply simpler, more natural punctuation. Corrected text favors periods and commas and avoids overusing semicolons or colons, while still staying free of em dashes and other non-standard punctuation.

Improved Text Formatting Modes

Built-in formatting modes received clearer instructions, resulting in noticeably smarter and safer rewrites across the board:

• Basic (grammar fixing) – Now treats everything you paste as text to correct, never as something to reply to or execute. It also safeguards subject matter and overall tone, leaves swearing untouched, sticks to plain ASCII punctuation with no em dashes or stray hyphens, and echoes back your text unchanged whenever nothing needs correcting—no fabricated "helpful" replies anymore.
• Casual mode – Rewrites questions, requests, and half-thought-out fragments into chatty phrasing rather than accidentally trying to respond to them. Profanity stays put where you put it, unnamed people remain anonymous, and placeholder brackets are kept intact, while still respecting an explicit wish to skip certain kinds of edits.
• Email mode – Less tempted to drift toward flowery prose, diplomatic swaps like "planning → excited", "goals → nice idea!", urges for speedy turnarounds, thank-yous out of nowhere, tagging coworkers, confetti/markdown in greetings, recruiting help mid-signature, steering titles, demanding open-ended expansions such as stats/research/plan-vs-request inversions meant to the writer, ticking off itemized bullets just to pad word counts, composing fictional paragraph closings, dangling recipients past opening line summaries, rushed voice-change to first person narrators, swapping decision into TODO riddles rather than options->apology extensions en route.
• Proofread mode... wait actually: that entry belongs further down—but worth double-checking once page preview loads fully—here = roughly equivalent trigger reading ✓

Improvements

• Text formatted with the Casual, Basic, and Romantic writing styles now sticks to standard keyboard punctuation only — no more em dashes or other special characters slipping into your text, so it looks consistent everywhere it's pasted.

Bug Fixes

• The app no longer quits when its last window is closed — it now keeps running in the background with its tray icon, so hotkeys and the recording indicator keep working. To exit completely, use the Quit option in the tray menu.
• The recording status indicator now reliably appears above other windows the moment a recording starts, and it no longer steals focus from the app you're currently using (macOS).
• When Screen Recording access isn't granted yet, the app now takes you straight to the Screen Recording page in System Settings, replacing an unreliable system prompt — so permission can be granted in one step (macOS).

Improvements

• Text enhancement results are now checked before being used: when the AI declines to process your text, that refusal is detected and the output is dropped cleanly instead of refusal text ending up in your transcription.
• If the AI's reply comes back in an unexpected form, enhancement now reports a clear error instead of passing raw, possibly malformed text through, so you know exactly what happened rather than seeing garbled output.

Snippets page redesign

• Edit snippets right on the card. Instead of opening a separate dialog to add or edit a snippet, you can now type triggers and replacement text directly on each snippet card — everything saves automatically as you type.
• Easier trigger management. Press Enter to quickly add another trigger to a snippet, and press Backspace on an empty trigger to remove it. You can also add triggers with the "Add trigger" button.
• Clearer duplicate warnings. If a trigger you enter is already used by another snippet, you'll now see a helpful message telling you which snippet it belongs to, shown right where you're typing.
• Help built in. A small info icon next to each trigger explains how triggers work — including that they're not case-sensitive, can match inside words, and how to write advanced pattern triggers.
• Improved layout and animations. Snippet cards look cleaner, with smoother animations when adding, deleting, dragging, or editing. Drag and delete controls are more visible and easier to tap.
• Undo deletions. After deleting a snippet, a notification lets you restore it with one click.
• Simplified flow for new snippets. Creating your first snippet or a new one now starts with a simple blank card instead of a multi-step dialog.

Fixes and improvements

• Validation messages for advanced (pattern-based) triggers are more accurate, and error text no longer shifts the layout while you type.
• The "Add trigger" button now stays out of the way until any errors are resolved, and hover/focus states across the snippets page are more consistent.

Improvements

• On macOS, overlays now stay smooth and keep updating in real time even when the app isn't in focus — they no longer pause, freeze, or lag while you're working in another application.

Bug Fixes

• Fixed an issue where the on-screen overlay shown during recording and processing could fail to appear — or display the wrong status — when it opens after a session has already started. The overlay now checks the current session state as soon as it launches and immediately shows the correct status.

Improvements

• Improved troubleshooting: activity happening in the app's windows, including the overlay, is now recorded in the app's diagnostic logs, so issues that occur in the interface are captured too.
• Overlay log entries now include session details, making it easier to trace what happened during a specific recording session when sharing logs for support.

New Features

• Multiple shortcuts per profile: You can now assign more than one shortcut to a profile. Turn on "Enable multiple hotkeys" in Advanced settings, then use the new "Add new" button in a profile's shortcut settings to add extra shortcuts for push-to-talk or toggle. Any of them activates the profile — handy if you use different keyboards or one key combination is taken by another app. With the option off, everything works like before with a single shortcut.
• Profile cards now display all of a profile's shortcuts side by side, joined with "or", instead of just one.
• The profile list on the account page now shows both the push-to-talk and toggle shortcuts, instead of only the push-to-talk one.

Improvements

• Existing profiles with a single shortcut carry over automatically — nothing to change or re-set on your end.
• The built-in health check now also flags invalid toggle shortcuts (previously it only checked push-to-talk ones), and its messages now tell you whether the problem is with a hold or toggle shortcut, making issues easier to locate and fix.

New Features

• Automatic history cleanup: Your history and its saved recordings can now be cleaned up automatically so they don't pile up forever. In Settings, a new "Data retention" section lets you choose how long to keep them — 7 days, 30 days (default), 90 days, 180 days, 1 year, or Never.
• The cleanup runs quietly on its own when the app starts — at most once a day — so you never have to think about it.

Behavior Changes

• History is now kept for 30 days by default. Anything older, along with its recording, will be automatically deleted the next time cleanup runs.
• If you want to keep your history and recordings indefinitely, go to Settings → Data retention and choose "Never" before upgrading, or as soon as possible after.

Improvements

• Data points on the overview chart are now narrower, giving the chart a cleaner, less cluttered look.

Improvements

• The activity chart on the Overview page now adapts to its size: when you're viewing a long time range or have a narrow window, nearby days are automatically grouped into evenly sized totals instead of showing one point per day, keeping the chart readable at any width.
• Hovering over a point on the Overview chart now shows the exact day (or date range) it covers, and the chart header displays the grouping currently in use — for example "Daily" or "4-day totals".
• Chart lines now make fuller use of the chart's height by scaling between the highest and lowest values in the selected range, making trends and day-to-day differences easier to spot.
• Polished the Overview page's loading placeholders so the layout stays steady and nothing overflows while your stats are loading.

macOS

• In Settings, macOS users now see a disabled Apple-logo button in place of "Check for Updates", since checking for updates isn't available on macOS.

Improvements

• More reliable text enhancement: When the AI assistant can't or won't process your text, the app now falls back to your original transcription instead of showing a refusal message — so you always get usable output.
• Cleaner results: If the enhancement step returns nothing meaningful, the app now discards it rather than inserting empty or broken text into your result.