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

v1.3.1

fix: UTF-8 encoding for Farsi/Arabic transcription on Windows systems with non-UTF-8 default locale

v1.3.0 - Settings UI, Google Cloud Engine, MIC Mode

What's new in v1.3.0

Settings Window

• Full dark-theme Settings UI — no more editing .env by hand
• All options configurable at runtime; settings saved to settings.json
• Engine-specific config (API key / model) expands inline below selection
• Changes apply instantly on Save

New Engine: Google Cloud STT

• Official Google Cloud Speech-to-Text REST API
• Requires a free API key from Google Cloud Console
• Higher accuracy than the unofficial engine

MIC Mode

• Always-on: mic stays open with a 500ms pre-roll buffer — first syllable never cut off
• On-demand: mic opens only while hotkey is held — more privacy
• Toggle live from tray icon menu — no restart needed

Update Checker

• Notifies via tray tooltip when a new version is available
• Can be disabled in Settings → Options

---

Engines

Engine	Key needed	Notes
Google (unofficial)	No	Free, no setup, great for Persian / Farsi
Google Cloud	Yes (free tier)	Official REST API, higher accuracy
Groq Whisper	Yes (free)	Fast, accurate, multilingual
Chrome Extension	No	Requires Chrome running in background

Download

Download SpeakPaste.exe below — single file, no install, runs directly.

First run: right-click tray icon → Settings to pick your engine and configure.

v1.1.0 - Unicode Typing Fix

What's New

• Fixed Unicode typing to work regardless of keyboard layout (EN/FA/any)
• Text now types correctly even when keyboard is set to English
• No clipboard interference
• Proper 64-bit Windows compatibility

Technical Details

• Switched from pynput to ctypes SendInput with KEYEVENTF_UNICODE
• Proper memory alignment for 64-bit systems
• Direct Unicode character injection (wVk=0, wScan=ord(char))

Download

Download SpeakPaste.exe below and place your .env file next to it with your GROQ_API_KEY.

SpeakPaste v1.0.0 - Voice to Text

Fix GitHub links in README