Skip to content

Releases: JMTDI/tt9

v59.0.1 futo

20 Jan 12:25
@JMTDI JMTDI
v59.0.1
ef763a0

Choose a tag to compare

View all tags
v59.0.1 futo Latest
Latest
  • Fixed auto stop listening for voice input.
  • Added latest commits @sspanak/tt9.
  • Changed apk signing certificate.
Assets 4
Loading

v59.0 Futo

10 Jan 15:22
@JMTDI JMTDI
v59.0
5f89225

Choose a tag to compare

View all tags
v59.0 Futo

TT9 v59.0 - Offline Voice Input Integration

🎤 Major Feature: Offline Voice Recognition

This release replaces TT9's online voice input with FUTO's offline Whisper-based speech-to-text engine, providing complete privacy and functionality without internet connectivity.

✨ Key Features

  • 100% Offline Operation: Voice recognition works entirely on-device with no data sent to external servers
  • 74+ Languages Supported: Automatic language detection and support for all major languages including:
    • English, Spanish, French, German, Italian, Portuguese, Russian, Chinese, Japanese, Korean, Arabic, Hindi
    • And 60+ more languages with automatic model downloading
  • Intelligent Model Management:
    • Built-in English model (43MB) for instant English recognition
    • Automatic multilingual model download (39MB) when using non-English languages
    • Smart model caching and switching
  • Privacy-Focused: All speech processing happens locally on your device
  • Voice Activity Detection: Automatic silence detection for natural voice input flow
  • Bluetooth Microphone Support: Seamlessly works with Bluetooth headsets and microphones

📦 What's Included

Lite Version (46.61 MB):

  • Complete offline voice input functionality
  • English model built-in
  • Multilingual model downloads on-demand
  • No dictionary data (text prediction available via online download)

Full Version (248.05 MB):

  • Everything in Lite version
  • All 44+ language dictionaries pre-installed
  • Complete offline text prediction and autocorrection
  • No internet required for any features

🛠️ Technical Details

  • Engine: whisper.cpp with GGML quantized models
  • Audio Processing: 16kHz mono PCM with VAD-based recording
  • Native Performance: JNI integration for efficient C++ inference
  • Language Code Mappings: Automatic handling of deprecated/variant codes (Indonesian, Hebrew, Yiddish, Norwegian variants, etc.)

🐛 Bug Fixes

  • Fixed composing text replacement issue when using voice input consecutively
  • Resolved JNI crashes during model loading
  • Fixed language detection for Bahasa Indonesia and other variant codes
  • Improved model download reliability with dual SSL certificate support

⚠️ Installation Notes

These are unsigned release builds. To install:

  1. Enable "Install from Unknown Sources" in Android settings
  2. You may need to sign the APK manually or use the debug builds for testing

Note: This release requires Android 5.0 (API 21) or higher. Voice input requires RECORD_AUDIO permission.

Loading