Release Notes

What's new in TalkKeys

v1.14.0 Latest June 2026

Real-time voice & the Signal widget

⚡ New 📊 Redesigned
⚡

Push-to-talk, in real time

Transcription now streams live as you speak instead of waiting until you finish. Press to talk, release to paste.

  • Connects the moment you press — first words land faster
  • Release to paste the finished text where you're typing
  • Press Esc anytime to cancel without pasting
📊

The new Signal widget

The floating widget is now a compact status strip that shows exactly what's happening during a session.

  • Live connection dot and a running transcript preview
  • Session timer plus a "too quiet" warning when your mic is low
  • One-tap retry if a session drops
🛟

More reliable sessions

A transcription error no longer disables dictation until you restart the app — the session recovers on its own.

  • Errors are now a recoverable state, not a dead end
  • Keep dictating right after a hiccup
v1.8.0 May 2026

Explain redesign

💡 Renamed 📩 New
💡

WTF → Explain

Same hotkey (Ctrl+Win+E), cleaner identity. Default popup is a single, professional explanation. The snarky "casual take" is now opt-in.

  • New name and icon — your existing settings still work
  • Settings → Explain → "Show casual take" to bring it back
  • Choose Savage or Sarcastic style when casual is on
📩

Smart Actions that actually work

The action buttons in the popup now do something. Click any of them after triggering Explain to transform the selected text.

  • Summarize, Simplify, Forward, Explain Code, Add Comments
  • One click — result inline with a Copy button
  • Powered by openai/gpt-oss-120b
✏️

Typed Reply (no more voice)

Click Reply on a selected message and type a short instruction (e.g., "say yes politely; mention I'll send Friday"). Get a properly-formatted reply back.

  • Email context → email-formatted reply (greeting + body + sign-off)
  • Chat context → short inline reply
  • Reply respects your Tone Rewriter setting
🪄

Cleaner output everywhere

All Explain outputs go through AI-marker stripping — no rogue em dashes, smart quotes, or invisible Unicode noise sneaking into your replies.

  • Em/en dashes → plain hyphens
  • Curly quotes → straight quotes
  • Ellipsis (…) → three dots; NBSP/ZWSP stripped
v1.7.0 May 2026

Tone for Dictation

🗣️ Opt-in 🛡️ Safe
🎚️

One Toggle, Whole-Pipeline Tone

Turn on a new checkbox in Settings → Tone and every dictation comes out in your configured tone. The same tone you set for the standalone Tone Rewriter (Ctrl+Win+R) is reused — nothing else to configure.

  • Settings → Tone → 'Also apply this tone to transcriptions'
  • Reuses your existing tone preset or custom prompt
  • Default off — existing users see no change until they opt in
🔁

Clean First, Then Tone

Your dictation still runs through smart cleaning (with your Words List intact). Once cleaned, it's passed through the Tone Rewriter and pasted in your voice.

  • Cleaning preserves your custom Words List
  • Tone applied before paste, not after
  • AI-style punctuation stripped automatically
🛡️

Failure-Safe

If the tone step times out, hits a network error, or returns nothing, the cleaned text still pastes. You never lose a dictation to a cosmetic step.

  • Network error → cleaned text pastes
  • Over-long text → cleaned text pastes
  • Fallback reason logged for diagnostics
v1.6.1 May 2026

Tone Presets

✨ New 🪄 Polish
✨

Four Tone Presets

One click to set the tone of every rewrite. Each preset preserves your meaning and structure - only the register shifts.

  • Friendly - warm, cooperative
  • Plain - direct, professional
  • Diplomatic - polite, tactful
  • Confident - assertive, no hedging
👁️

See Before You Pick

Settings shows a sample input vs. output pair for the active preset, so you can compare styles at a glance.

  • Shared sample input across all presets
  • Per-preset 'after' updates on click
  • Your custom prompt still works
🪄

Cleaner Output

Rewrites no longer scream 'I was written by AI.' Em dashes, smart quotes, ellipsis characters, and other punctuation tells are stripped automatically.

  • Em dashes become space-hyphen-space
  • Smart quotes become straight quotes
  • Always-on, no setting to toggle
v1.6.0 May 2026

Tone Rewriter

✍️ New 🛡️ Safety
✍️

Tone Rewriter

Select text in any app, press Ctrl+Win+R, and the selection is rewritten in your configured tone - pasted right back where the cursor was.

  • Press Ctrl+Win+R from any app
  • Default tone: friendly and cooperative
  • Powered by openai/gpt-oss-120b
⚙️

Customize the Tone

Edit the tone prompt in Settings to match your voice - formal, terse, technical, whatever fits.

  • Multi-line prompt editor in Settings
  • Changes apply instantly, no restart
  • Empty prompt? Default is restored
🛡️

Never Lose a Rewrite

If the paste cannot land (read-only field, focus changed), a fallback dialog shows the rewrite and keeps it on your clipboard.

  • One-click Copy button in fallback dialog
  • Original clipboard restored on success
  • Hotkey conflict detection in Settings
v1.4.0 December 2025

Premium Subscriptions

⭐ $5/mo 📊 Real-time
⭐

Premium Subscription

Free users get 10 minutes per day. Upgrade to Premium for 60 minutes of daily transcription - perfect for professionals and power users.

  • 60 minutes daily (6x more than free)
  • $5/month - cancel anytime
  • Manage subscription at talkkeys.symphonytek.dk/account
📊

Usage Dashboard

Your subscription tier and daily usage are now visible right in Settings. Track how much time you've used today.

  • FREE or PREMIUM badge display
  • Visual usage progress bar
  • Quick link to manage account
🔔

Limit Notifications

When you reach your daily limit, you'll get a helpful notification with an easy upgrade option.

  • Toast notification at limit
  • One-click upgrade button
  • Automatic sync after recording
v1.3.1 December 2025

Settings Bug Fixes

🔧 Fixed 🛡️ Stable
🔧

Account Card Fixed

The account section in Settings now updates immediately when you sign in or sign out - no more stale data.

  • Sign-in immediately reflects in UI
  • Sign-out clears account display
  • No restart required
🛡️

Settings Stability

Fixed MVVM data binding issues that caused some Settings controls to not respond properly.

  • All buttons now respond correctly
  • Improved ViewModel-View sync
  • More reliable Save & Close
v1.3.0 December 2025

Audio Latency Fix & MVVM Architecture

🎙️ Fixed 🏗️ Internal
🎙️

Recording Latency Fixed

Recording now waits until the microphone is truly ready before prompting you to speak. No more missing your first word!

  • 'Getting ready...' during mic initialization
  • Prominent green 'Speak now!' when ready
  • Clear visual feedback
⚙️

Microphone Selection

Settings microphone selection now uses a proper dialog with radio buttons for easier selection.

  • Clean selection dialog
  • Radio buttons for each device
  • Instant device switching
🏗️

Clean Architecture

Internal restructuring to 3-project MVVM architecture. This enables future cross-platform support (Mac/Linux).

  • TalkKeys.Core - Platform-agnostic logic
  • TalkKeys.ViewModels - MVVM layer
  • TalkKeys - WPF UI layer