v1.13.0

What's new

• Remote Desktop hotkey handover. Run SpeakPaste on your PC and inside the remote session with the same hotkey. While a Remote Desktop window (mstsc, msrdc, vmconnect, rdcman) is in front, the local copy leaves the hotkey to the remote one. Setting: General → Leave hotkeys to the remote PC while a Remote Desktop window is in front (on by default). Setup notes in the README.
• Self-healing microphone. When the input device disappears (a Remote Desktop reconnect, a USB mic unplugged, a new default device), SpeakPaste reopens it within a few seconds. No more restarting after a reconnect.
• Live History. New voices show up while the History window is open; before, the window stopped updating once history was full.

Fixes

• The version string was stuck at 1.2.0, so every start reported an update to 1.12.0.
• The log names the focused app on every hotkey press, for diagnosing where a press went.

v1.12.0

What's new in v1.12.0

The hotkey no longer gets stuck

• A hotkey could silently degrade: after a busy moment, pressing Alt alone started recording, as if Win were still held. Pressing the real hotkey once cleared it, until it happened again.
• Key state now comes straight from Windows instead of from tracked keyboard events, so a missed key release can no longer wedge it. The same fix removes a case where typing could hang forever.

Use your own Gemini endpoint

• New Base URL and Base URL token fields on the Speech → Text tab. Point them at any Gemini-compatible proxy and every Gemini call - transcription, prompt rewriting and text-to-speech - goes there instead of to Google, with your real key staying on the proxy.
• Environment variables SPEAKPASTE_GEMINI_BASE_URL and SPEAKPASTE_GEMINI_TOKEN override the fields.
• Leave them empty and nothing changes.

v1.11.0

What's new in v1.11.0

Hotkeys

• A Record button next to each hotkey field captures the combination you press instead of making you type it.
• Combinations are validated. Windows-reserved ones are rejected with the reason - including that a Win+letter combo stays reserved even with extra modifiers held, so a broken hotkey is caught before it ever gets saved.

Text insertion

• New Automatic / Always type / Always paste setting. Automatic types short text and pastes anything over 220 characters, where typing key by key would take seconds. Pasting restores your previous clipboard content afterwards.
• Gives apps that ignore synthetic keystrokes (SendInput) a working path.

Error reporting

• A failed transcription now shows a small notification with the actual reason instead of only writing a line to a log nobody sees.

Player widget

• The floating reading player is now about 40% shorter: the speed control moved onto the same row as play/pause/next/prev/stop, leaving only the timeline on its own row below.

Tests

• 34 automated tests now run on every push, covering hotkey validation, text chunking, playback speed accuracy, and more.

v1.10.0

What's new in v1.10.0

Read any text aloud

• Select text in any Windows app and press Win+Shift to hear it read.
• Your clipboard is untouched — the selection is copied and your previous clipboard content is put back.
• A floating player gives you play/pause, stop, a seek bar, and a 0.5x–2x speed slider that keeps the voice natural instead of turning it into a chipmunk.
• Long text starts playing within a few seconds instead of making you wait: it is split at sentence boundaries and later parts are prepared while earlier ones play, with no gaps between them.
• Jump between parts with the ◀◀ / ▶▶ buttons. A part 2/5 counter and tick marks on the seek bar show where you are. Pressing back mid-part replays that part from its start.
• Two voice engines: free Microsoft Edge neural voices (default, no key), or Gemini-TTS on Vertex AI where you can describe in plain words how it should read.
• The tray menu also has Read clipboard aloud and Stop reading.

New gemini speech-to-text engine

• Detects the spoken language by itself, so the keyboard layout and the Language setting no longer matter.
• Persian sentences keep their English technical terms in Latin script — say "برو روی branch جدید و commit رو merge کن" and branch, commit and merge come out in English instead of being mangled.

Settings and history

• Settings are now split into three tabs: Speech → Text, Text → Speech, and General.
• A Test voice button plays a sample without closing the window.
• Click any history entry to copy it back to your clipboard.

Housekeeping

• SpeakPaste is now licensed under Apache 2.0. A THIRD-PARTY-NOTICES.md documents every dependency licence and the LGPL/GPL components bundled into the executable.
• A rotating log file is written next to the app, which makes problems diagnosable now that the app runs without a console.

v1.9.0

Fixes and improvements:

• Settings and History windows no longer hang or fail to open
• Fix Google engine [WinError 2] in the packaged exe (bundled FLAC encoder)
• History right-aligns Persian text, keeps English left-aligned
• History now persists across restarts
• Fix Google Cloud API key not applying on Save

v1.8.0

• Gemini thinking level + media resolution controls (Minimal/Low/Medium/High), default Low — reduces latency\n- Settings window is now scrollable\n- Log lines now show timestamp and elapsed time per step

v1.7.0

What's new in v1.7.0

Decoupled Transcription Engine from Prompt Mode

• Transcription Engine and Prompt are now two independent settings
• Choose any STT engine (Google, Groq, Google Cloud, Chrome Extension) independently
• Choose a prompt mode (Off, Gemini Flash Lite, Gemini Flash) independently
• Selecting Gemini Flash (voice to prompt directly) grays out the Transcription Engine section - it is bypassed entirely
• Settings auto-migrate from the old combined engine format

v1.6.0

What's new in v1.6.0

History window

• New History item in the tray right-click menu
• Shows all transcriptions from the current session with timestamp and engine
• Updates in real-time — new entries appear instantly while the window is open
• For gemini-lite (two-step processing): shows both the raw voice transcription and the converted English prompt as separate labeled rows
• Show voice text checkbox to toggle the raw STT row for gemini-lite entries
• Clear button to wipe the session history
• History is in-memory only and resets on restart

v1.5.0

What's new in v1.5.0

Auto-detect language from Windows keyboard layout

• New "Follow Windows keyboard layout" option in Settings → General
• When enabled, SpeakPaste reads your active keyboard layout at the moment you press the hotkey - no manual language switching needed
• Toggle Persian layout (Alt+Shift) → speak Persian; switch to English layout → speak English
• Language is locked for the full press → release → transcribe cycle
• Falls back to the configured language if the layout is not recognised
• Supported layouts: Persian/Farsi, English, Arabic, Turkish, German, French, Russian, Portuguese, Spanish, Japanese, Korean, Chinese
• Works with all engines: google, google-cloud, groq, google-ext, gemini-lite

v1.4.0

What's new in v1.4.0

Gemini Engines

• gemini-lite — speak in any language → Google STT (free) → Gemini Flash Lite → clean English programming prompt for your AI coding assistant
• gemini-flash — speak in any language → WAV sent directly to Gemini Flash (multimodal) → English prompt, no intermediate STT step
• System prompt is fully customizable in Settings
• Free Gemini API key from aistudio.google.com

CI/CD Pipeline

• GitHub Actions workflow: push a tag → exe built automatically on Windows runner → published as GitHub Release
• release.py helper script for local version bump + tag + push