Weesper Neon Flow

Changelog

Weesper Neon Flow changelog: every release. New features, bug fixes, performance improvements, and AI model updates for Mac and Windows. Updated each version.

Version 1.7.9

✨ New Features

  • Turn on GPU acceleration yourself — a new opt-in setting lets you enable graphics-card acceleration for faster dictation on supported Windows PCs, with an automatic fallback to your processor if the hardware is not compatible.
  • GPU acceleration for Intel Macs — Intel-based Macs can now opt in to Metal acceleration for quicker transcription.

🚀 Improvements

  • Leaner, faster speech engine — dictation now runs entirely in-process, trimming startup time and shrinking the app.
  • More resilient first run — a built-in fallback model keeps dictation working even before a full model finishes setting up.

🐛 Bug Fixes

  • Fixed a rare startup crash on some Macs during speech-engine initialization.
  • More reliable precision translation — fixed an issue that could block translation in Precision mode.

Version 1.7.8

✨ New Features

  • GPU-accelerated dictation on Windows — Weesper now ships with Vulkan GPU acceleration built in, using your graphics card for faster, smoother transcription on compatible PCs.

🚀 Improvements

  • Optimized for your processor on Windows — the app now automatically loads the speech engine build best matched to your CPU for better speed and reliability.
  • Smart profile suggestions — if dictation runs slowly on your machine, Weesper now recommends a lighter, faster speech profile.
  • Clearer speech model names — the "Turbo" profile is now called "Precision".
  • Refreshed home screen — a cleaner two-column layout with more consistent window sizing.

🐛 Bug Fixes

  • Fixed a startup crash affecting some Windows PCs.
  • Fixed the main window opening partly off-screen on small displays.
  • Fewer false timeouts during transcription on Windows.
  • Correct app icon now appears on the Windows desktop shortcut.

Version 1.7.7

🚀 Improvements

  • More stable Windows dictation — refined CPU profile for smoother, more reliable transcription on Intel and AMD machines.
  • Smarter default language — dictation now defaults to your operating system language on first launch, reducing manual setup.
  • Unified transcription engine — consolidated the underlying speech runtime for more consistent results across macOS and Windows.

🐛 Bug Fixes

  • Speech model selection now persists — your chosen profile is correctly remembered between sessions.

Version 1.7.6

✨ New Features

  • New press-to-toggle dictation mode — switch your recording shortcut between the default hold-to-speak behavior and a new press-to-start/press-to-stop mode for longer, hands-free sessions. Configure it in Settings → Audio Setup → Recording Shortcut.

🐛 Bug Fixes

  • More accurate auto-language detection — fixed a regression where the speech engine could pick the wrong language on short or noisy utterances.

Version 1.7.5

✨ New Features

  • Redesigned dictionary — manage your custom snippets faster with full keyboard navigation, instant Backspace-to-delete, and a polished card layout.
  • Enhanced history view — browse your past transcriptions with a sleeker interface, jump between entries with the keyboard, and re-insert any past dictation directly into your active app.
  • Smarter shortcuts management — text shortcuts now ship with example presets, time markers (insert current date/time), and a refined import/export flow.
  • Native speaker diarization in Meeting mode — Weesper now automatically identifies and labels multiple speakers during your meeting recordings.
  • Small speech profiles are back — the small and small.en models return as a lightweight option for short dictations on lower-spec machines.
  • Keyboard navigation in Settings — move through every configuration card without touching the mouse.

🚀 Improvements

  • Faster dictation feedback — the HUD now appears almost instantly when you press your hotkey, making short dictations feel snappier.
  • Better performance on Intel Macs — older Intel Macs now use an optimized transcription path for more reliable dictation.
  • Optional Vulkan GPU acceleration on Windows — the rock-solid CPU path remains the default for production, while Vulkan is available as an opt-in experiment for compatible GPUs.
  • More resilient audio capture — the recording pipeline now recovers gracefully from device hiccups and silent gaps without dropping your session.
  • Streamlined transcription engine — faster startup, smoother decoding, and more accurate auto-language detection on short utterances.

🐛 Bug Fixes

  • Fewer dropped characters when injecting long dictations into target applications.
  • Fixed Meeting hotkey that occasionally failed to start recording.
  • Fixed meeting recorder stuck-stream issue that could leave a recording hanging.
  • Fixed dictation recovery when the active app loses focus mid-injection.
  • Fixed window flash on startup on Mac — the main window no longer briefly appears when launching minimized.
  • Fixed several auto-language detection mistakes on short or noisy utterances.

Version 1.7.4

🚀 Improvements

  • More reliable Windows transcription — Weesper now uses a hardened CPU path on Windows by default, with significantly fewer transcription failures across older and lower-spec machines.
  • Self-contained Windows runtime — the app now ships with its speech engine fully embedded, removing a class of runtime errors that could cause silent failures after install or update.

🐛 Bug Fixes

  • Fixed an elevated Windows failure rate observed shortly after the 1.7.3 rollout on a subset of PCs — transcription completion rates are back to expected levels.
  • Fixed a Windows crash during startup caused by missing runtime components on some configurations.
  • Improved release-time safety checks — additional automated guards now run on every Windows build to catch regressions before they reach you.

Version 1.7.3

✨ New Features

  • Vulkan GPU acceleration on Windows — modern PCs with a compatible graphics card now get noticeably faster transcription out of the box, automatically.
  • Refreshed speech models catalog — updated model lineup with safer Turbo defaults for high-end machines and clearer recommendations for your hardware.

🚀 Improvements

  • Faster Windows transcription engine — switched to a tighter, more efficient core pipeline; you will feel the difference between releasing the hotkey and seeing your text appear.
  • Smarter engine routing — Weesper now picks the best transcription path (GPU or CPU) automatically based on your hardware.
  • More accurate progress feedback — the processing indicator now reflects real warm-up and decoding steps so long transcriptions feel less mysterious.
  • Persistent Free quota banner — when you approach your trial limit, the indicator stays visible with a one-tap upgrade.

🐛 Bug Fixes

  • Fixed a crash on Windows machines that did not expose a detected GPU to the transcription engine.
  • Fixed a rare HUD freeze after a silent dictation — focus is now properly returned to your previous app.
  • Fixed a silent startup failure where the dictation engine could fail to initialize without notice.
  • Fixed an auto-insertion edge case on Windows affecting some target applications.

Version 1.7.2

✨ New Features

  • Whisper model RAM retention — control how long the speech model stays loaded between dictations: release immediately (Auto), hold for a configurable timeout, or keep it in memory permanently (Always) for instant response on every use.

🚀 Improvements

  • Windows: accurate GPU backend reporting — Weesper now correctly identifies and reports Vulkan as the active acceleration backend.
  • Windows: tuned transcription thread policy — better core utilization during speech recognition on multi-core systems.
  • Cleaner transcriptions — removed an unreliable enumeration auto-formatting heuristic that could unexpectedly reformat dictated text.

🐛 Bug Fixes

  • Fixed a rare Windows crash when the app window was closed during an active dictation session.
  • macOS: eliminated Metal false positives — the GPU health check no longer incorrectly flags a working Metal setup as failed.
  • Fixed trial expiry emails — the trial end date now appears as a readable date instead of a raw ISO timestamp.
  • Removed a duplicate troubleshooting link from the Quality settings panel.

Version 1.7.1

🚀 Improvements

  • Fixed Windows dictation latency regression — transcription now starts as quickly as before, restoring the snappy response time you expect.
  • More resilient voice activity detection on Windows — the VAD model now deploys reliably even when the install directory is restricted.
  • Crash-resistant settings storage — state files are now written atomically so they cannot be corrupted if Weesper is interrupted unexpectedly.

🐛 Bug Fixes

  • macOS Intel stability — when the Metal GPU backend cannot start, Weesper now falls back to CPU automatically and recovers cleanly after a crash.
  • Clearer fallback notifications — when speech recognition has to switch to a safer mode, a notification now explains what happened.

Version 1.7.0

✨ New Features

  • Adaptive CPU backend on Windows — Weesper now picks the fastest CPU instruction set available on your machine (Haswell baseline up to Ice Lake), reducing inference latency on modern CPUs while keeping compatibility with all AVX2 hardware.

🚀 Improvements

  • Smarter first-launch defaults — speech model, GPU acceleration toggle, and dictation quality preset now use OS-specific defaults so a fresh install feels right out of the box.
  • Streamlined sign-up — EULA consent and email entry are now combined in a single onboarding step.
  • Faster Windows dictation with large models — the thread budget now uses every available core when transcribing with bigger models.
  • Reduced dictation startup delay — the microphone audio stream stays warm between sessions.
  • New Help recipes for diagnosing slow or low-quality dictation.

🐛 Bug Fixes

  • Robust CPU backend fallback — if Weesper's adaptive CPU backend cannot start, the app now stays running and reports a clear error instead of crashing.

Version 1.6.2

🚀 Improvements

  • Windows installer is now fully offline — bundles the WebView2 runtime and the Microsoft Visual C++ runtime, so installs work on fresh machines without internet access.
  • Windows installer no longer triggers Microsoft Defender or antivirus false positives, ensuring a smoother first-time setup.
  • Direct and Meeting Flow installs are now fully isolated — installing one variant no longer interferes with the other.

🐛 Bug Fixes

  • Your subscription email is now preserved when you re-link your account.
  • Windows: helper process console windows no longer flash on screen during dictation.

Version 1.6.1

🐛 Bug Fixes

  • Fixed a Windows crash during transcription on older Intel CPUs (5th and 6th generation Broadwell/Skylake — e.g. Lenovo T450). The app would previously quit silently when starting a transcription. AVX2-capable processors are now fully supported as originally intended.

Version 1.6.0

✨ New Features

  • GPU Acceleration Toggle — A new switch in Settings lets you turn GPU acceleration on or off on Windows and macOS, so you can choose between maximum speed or saving battery on the fly.
  • Trailing Punctuation Modes — Choose how dictation ends each transcription: full automatic punctuation, just a final period, or none at all. Perfect for chat, code, or formal writing.
  • Free Lite Mode After Trial — When your free trial ends, you can keep dictating short clips up to 60 seconds for free, with clear in-HUD feedback when the cap is reached.
  • Audio Cues Preference — Toggle the subtle start/stop sounds that confirm dictation is recording.
  • Smarter Speed Guidance — When you pick a heavy quality preset on a slower Windows machine, Weesper now suggests the best preset for your hardware.

🚀 Improvements

  • Faster Windows Dictation — Windows GPU acceleration is now wired end-to-end through the speech engine, dramatically reducing transcription time on compatible PCs.
  • Renamed "Instant" Speed Preset to "Fast" — Clearer naming so the preset honestly describes what it delivers.
  • More Reliable Text Injection on Windows — Hardened the fallback so dictated text reaches apps that don't accept standard keyboard input (terminals, legacy software, secure fields).
  • Steadier HUD Voice Visualizer — Refreshed the audio meter for smoother, more accurate voice tracking while you speak.
  • Cleaner Transcriptions — Reinforced protection against prompt-instruction leaks so your dictation output stays free of stray system text.
  • Sharper Dictation & Meeting Stability — Hardened both pipelines for long, back-to-back, or interrupted sessions.
  • Refreshed UI Copy — Clearer wording across quality presets, GPU settings, recognition profiles, activation, and language selectors.
  • Stronger Privacy Defaults — Disabling analytics now immediately purges the local queue and snapshot, with no leftover events ever sent.
  • Better macOS Launch at Login — Migrated to the modern SMAppService framework for smoother startup on macOS 13 and later.
  • Smoother Lifetime Upgrade — Clearer diagnostics and recovery paths when upgrading from monthly to lifetime, plus automatic VAT invoices on every paid checkout.

🐛 Bug Fixes

  • Fixed a Windows crash that could occur when closing the app while dictation was active.
  • Fixed text not always being inserted on Windows when the target app required UI Automation.
  • Fixed the Windows system tray icon not restoring the main window on click.
  • Fixed the Settings panel losing keyboard focus when a dictation session started.
  • Fixed the Spacebar hotkey occasionally being mis-recognized after rapid switching.
  • Fixed a rare crash on older Intel Macs during transcription startup.
  • Fixed lifetime license upgrade not completing correctly in some Stripe wallet checkouts.
  • Fixed the activation window failing to open in some email-verification scenarios.
  • Fixed missing French accents in the auto-update settings copy.

Version 1.5.0

✨ New Features

  • Windows GPU acceleration — Transcription on Windows now runs on Vulkan-capable GPUs when available, with automatic detection at launch for dramatically faster processing.
  • French consent flow — The checkout experience now includes locale-aware consent acknowledgements to align with French privacy requirements.

🚀 Improvements

  • Cleaner transcriptions — Stronger protection against prompt echo in the speech-to-text pipeline for more reliable dictation output.

🐛 Bug Fixes

  • Fixed a rare crash during silence trimming on very short recordings.
  • Improved email delivery reliability for payment confirmations, trial notifications, and subscription events.
  • Updated bundled dependencies with the latest security patches.

Version 1.4.13

🚀 Improvements

  • Upgraded Speech Engine — Updated Whisper transcription stack with simplified GPU acceleration on macOS for faster and more accurate dictation
  • More Reliable Text Insertion — Hardened the text injection mechanism on both macOS and Windows, with better recovery when the target application is slow to respond
  • Updated Dependencies — Refreshed all third-party libraries to their latest secure versions

🐛 Bug Fixes

  • Fixed Windows version not being detected correctly in system information (was always showing "unknown")
  • Fixed activation window not opening when email verification is required
  • Fixed clipboard restoration on macOS after paste-based text insertion

Version 1.4.12

🚀 Improvements

  • Faster Dictation Start — Reduced latency when starting dictation, with speech engine pre-warming on Windows for near-instant readiness

🐛 Bug Fixes

  • Fixed clipboard paste fallback on macOS not handling special characters (escape sequences, accented letters) correctly

Version 1.4.11

🚀 Improvements

  • Online Help Center — The app now links directly to the comprehensive Help Center on the website, with guides and troubleshooting available in 6 languages
  • Dictionary Escape Sequences — Use \n and \t in custom dictionary written forms for line breaks and tabs
  • Clearer Tray Menu — "Universal Translator" renamed to "Languages" for better clarity

🐛 Bug Fixes

  • Fixed an inaccurate description of punctuation voice commands across all 13 languages

Version 1.4.10

🚀 Improvements

  • Faster Dictation for More Users — Improved speech engine fallback logic so more dictations use the fast VAD-accelerated path. Previously, up to 40% of dictations could fall back to a slower processing mode; the engine now recovers automatically when the voice detection model is temporarily unavailable
  • Smarter Short Audio Handling — Very short recordings (accidental hotkey taps) are now detected earlier, avoiding unnecessary processing and reducing wasted CPU cycles
  • More Reliable Licensing — Strengthened subscription and activation state synchronization. Lifetime purchases made as a guest are now correctly recognized, and license validation is more resilient to transient network issues

🐛 Bug Fixes

  • Fixed an edge case where subscription status could temporarily desync after switching between devices or restarting the app
  • Fixed an issue where lifetime purchases made without an existing Stripe customer account were not immediately activated
  • Improved crash diagnostics with structured classification, making it easier to identify and fix issues faster in future updates

Version 1.4.6

✨ New Features

  • New Minimal Profile — The smallest multilingual transcription engine (~75 MB). Makes dictation usable even on the least powerful machines
  • Full 13-Language Support — The app is now fully translated into 13 languages: German, Spanish, French, Italian, Japanese, Korean, Dutch, Polish, Brazilian Portuguese, Russian, Turkish, and Simplified Chinese
  • Faster & More Reliable Updates — Automatic updates are now delivered through a new global distribution network (Cloudflare R2) for faster downloads worldwide

🚀 Improvements

  • Better memory management and simultaneous recording handling for smoother dictation
  • More robust transcription pipeline when switching engines mid-use
  • Strengthened voice activity detection (VAD) with better edge-case handling

🐛 Bug Fixes

  • Fixed a bug where long recordings (40s+) with the Fn key would not stop on release, producing abnormally long transcriptions
  • Fixed audio buffer contamination when switching models, which could produce corrupted transcriptions
  • Fixed embedded transcription engine path in the packaged application

Version 1.4.5

✨ New Features

  • Auto-Detect Language — The Universal Translator now automatically detects the source language. No need to select it manually anymore
  • Hotkey Status Alerts — If another app conflicts with your shortcut, Weesper now alerts you instantly with a clear explanation and how to fix it
  • Context Prompts Auto-Save — Your context prompts are saved automatically as you type. No more manual saving
  • Flexible Plans & Lifetime Option — Redesigned plan picker with clearer pricing and a new lifetime purchase option

🚀 Improvements

  • Redesigned History Panel — Cleaner action buttons with expand/collapse for easier browsing
  • Faster Updates — App updates now delivered via a global CDN for faster, more reliable downloads
  • UI Polish — Refreshed navigation icons, refined sidebar alignment, and instant hotkey changes without restart

🐛 Bug Fixes

  • Fixed Windows fullscreen apps losing focus when restoring Weesper
  • Fixed hotkey labels not updating when switching language
  • Fixed visual glitch on selected option cards

Version 1.4.1

🚀 Improvements

  • Even More Stable — The app now recovers gracefully from more edge cases, including corrupted settings files or interrupted background tasks
  • Faster Future Updates — Internal architecture improvements that make the app easier to maintain and update going forward

🐛 Bug Fixes

  • Fixed timestamps leaking into dictation text — Voice transcription now inserts clean text without time markers
  • Fixed radio button borders being cut off in scrollable settings panels
  • Fixed About popup showing incorrect information on some platforms

Version 1.4.0

✨ New Features

  • Custom Dictionary — Add your own vocabulary words for better recognition accuracy, and enable auto-correct to fix common transcription mistakes automatically

🚀 Improvements

  • Simplified Navigation — Sidebar streamlined from 19 to 15 sections for faster access to what matters
  • Better Dark Mode Contrast — Improved readability for history entries and option cards in dark mode
  • Lower Memory Usage — Reduced memory pressure during long recording sessions
  • Crash-proof Engine — Comprehensive error-handling hardening across the entire backend — the app now gracefully recovers from unexpected states instead of crashing
  • Refreshed UI Theme — Polished visual consistency with a layered token-based design system

🐛 Bug Fixes

  • Fixed clipboard encoding for non-ASCII characters when copying from History on macOS
  • Fixed input language dropdown label being overwritten after selecting a language

Version 1.3.9

🚀 Improvements

  • More responsive interface — Save indicator now reacts to shortcut captures, Reset to Defaults button requires confirmation, and About window closes properly on Windows
  • Clearer Windows interface — Shortened menu labels prevent text truncation, improved notification icon with purple tooltip, and redesigned notifications that don't shift the page
  • Better model management — English profiles stay visible at all times with accessible headers and automatically reload when opening the relevant section
  • Simplified dictionary workflow — Your word additions and deletions are automatically saved without needing to click Save or Reset buttons
  • Real-time version display — The app now shows the current version in System Information for both Direct and Lite editions, ensuring accurate version tracking
  • Enhanced Lite edition — Improved text labels and shortcuts (e.g., "hashtag address"), with contact and system information aligned with actual runtime data
  • Overall polish — Various refinements including cleaner code, Unicode dialog support, and updated documentation for a more reliable experience

Version 1.3.7

✨ New Features

  • Time Saved statistics — Track your productivity gains with detailed breakdown showing typing time vs voice time and percentage saved
  • Changelog in sidebar — Click the version number to see what's new in each release

🚀 Improvements

  • Up to 65% faster transcription on Windows — Optimized performance for modern processors
  • Smart pause detection — Optional 20-30% speed boost by automatically skipping silent moments (Settings > Quality & Speed, experimental)
  • Better light mode readability — Improved contrast with darker text and clearer borders
  • Scrollable sidebar — Better interface adaptation when resizing the window
  • Enhanced visual design — Improved icons for better clarity and accessibility

🐛 Bug Fixes

  • Fixed critical macOS startup issue — The app now launches reliably without errors
  • More stable overlay display — Improved reliability prevents interface freezing
  • Fixed transcription history — Past dictations now display correctly

Version 1.3.0

🚀 Improvements

  • 30-50% faster transcription on Windows — Optimized performance for modern processors (Intel 2013+, AMD 2015+/Ryzen). Your dictations are now significantly faster
  • Smarter error messages — The app now clearly explains when a setting cannot be applied and why, preventing invalid configurations
  • Improved settings management — Valid settings are saved even if some changes cannot be applied

🐛 Bug Fixes

  • Prevented app from becoming inaccessible when disabling both menu bar and dock icons

Version 1.2.10

✨ New Features

  • Auto-update toggle directly accessible from app menu and synchronized with menu bar icon
  • System notifications when updates are detected or installed

🚀 Improvements

  • Optimized interface layout for smoother scrolling and better visual harmony
  • More reliable update system on macOS and Windows with seamless installation

Version 1.2.0

✨ New Features

  • Added contextual prompts for professional emails and technical documentation
  • New keyboard shortcut customization options

🚀 Improvements

  • 50% faster transcription on Apple Silicon (M1/M2/M3/M4)
  • Improved accuracy for technical terms and acronyms
  • Better punctuation placement in long sentences

🐛 Bug Fixes

  • Fixed clipboard insertion on Windows 11 24H2
  • Resolved microphone permission dialog appearing twice on macOS
  • Fixed rare crash when switching between apps during dictation

Version 1.1.0

✨ New Features

  • Windows 10/11 support now available
  • Dark mode UI improvements
  • Export transcription history to text file

🚀 Improvements

  • Reduced memory usage by 30%
  • Faster app launch time
  • Enhanced microphone noise filtering

Version 1.0.0

🎉 Initial Release

  • 100% offline voice dictation for macOS
  • Universal text insertion in any application
  • Support for 55+ languages including French, English, Spanish, German, Italian, Portuguese
  • Metal acceleration for Apple Silicon
  • 15-day free trial with no credit card required
Ready to supercharge your workflow? Download Free Trial