Live Translator Overlay

A high-performance, real-time subtitle and OCR translation overlay for Windows.
Translate live game audio, video streams, voice meetings, and on-screen text directly on top of your display with zero click-through disruption.

---

Features

• 100% Free and Unlimited by Default (No API Key Required)
  • Operates out-of-the-box using Google Translate's high-speed endpoint (~100–200ms latency).
  • No account creation, no credit card, and no token quotas.
• Real-Time Speech-to-Text (STT)
  • Deepgram Streaming (Recommended): Sub-second latency WebSocket streaming with live word-by-word interim preview and multi-speaker diarization with distinct color identification.
  • Groq Whisper (Large-v3): Cloud batch transcription powered by Whisper Large-v3.
• Anti-Hallucination and Voice Activity Detection (VAD)
  • Audio energy gating prevents ambient noise or silence from generating spurious transcriptions.
  • Multi-language filtering actively blocks common subtitle artifacts (such as "Subtitles by...", "Spasiba", "To be continued...", etc.).
• Live Screen OCR Translation
  • Draw any bounding rectangle to translate foreign in-game dialogues, manga, or application interfaces in real time using native Windows Hardware-Accelerated OCR.
• Non-Intrusive Desktop Overlay
  • Always-on-top transparent canvas: Captions render cleanly above full-screen games, media players, and desktop applications.
  • Freely repositionable: Move the subtitle banner and settings dialog anywhere on your monitors via drag-and-drop.
  • Screen Capture Exclusion (Privacy Mode): Employs Windows Graphics Capture Exclusion (SetWindowDisplayAffinity) to keep captions and all application windows completely hidden from OBS Studio, Discord screen sharing, Zoom, and screen recordings, with built-in Snipping Tool deadlock protection.
  • Translation History: Integrated history log to review past spoken and translated statements with AI-powered conversation summaries.
• Silent Background Auto-Updater
  • Starting from v1.2.0, the app checks for new releases on GitHub and downloads updates silently in the background.
  • Updates are applied automatically on application restart or exit while strictly preserving user API keys and configurations.
• Multi-Language User Interface & Target Translation
  • Fully localized in English, Indonesian (Bahasa Indonesia), and Russian (Русский).
  • Automatically matches your Windows display language by default, with dynamic in-app language switching and auto target language detection.
• Instant Groq AI Key Verification & Engine Auto-Switching
  • Live API key verification with real-time connection feedback and automatic model population (powered by llama-3.3-70b-versatile).
• In-App Bug Reporting & Resolution Tracker
  • Report bugs, feature suggestions, and criticism directly inside the application without needing third-party accounts.
  • My Reports History: Track previously submitted reports and monitor resolution progress (Pending, In Progress, Fixed) per device.
  • Automatic status resolution sync when newer application updates are released.

---

Quick Start

Option 1: Pre-Built Binary (Recommended)

1. Navigate to the Releases page.
2. Download the latest LiveTranslatorOverlay-Windows-x64.zip.
3. Extract the ZIP archive to your desired location.
4. Launch LiveTranslatorOverlay.exe.
5. (Optional) If prompted, install the .NET 10 Desktop Runtime.

Option 2: Build From Source

Ensure that the .NET 10 SDK is installed on your machine.

# Clone the repository
git clone https://github.com/nnavyy/LiveTranslatorOverlay.git
cd LiveTranslatorOverlay

# Restore and compile release build
dotnet build -c Release

# Execute application
dotnet run -c Release

---

Configuration and API Setup (Bring Your Own Key)

The application operates without charge using Google Translate. If you prefer to utilize specialized cloud AI models, you may optionally supply your own API keys.

Service	Category	Pricing	Intended Use Case
Google Translate	Translation	Free (No Key Required)	Daily translation, streaming, and gaming without rate limits
Deepgram	Speech-to-Text	Free Tier ($200 credit / ~12,000 mins)	Low-latency streaming transcription and speaker separation
Groq AI	Translation & Whisper STT	Free Tier (Rate-limited tokens)	Context-aware LLM translations and Whisper Large-v3

Supplying API Keys (Optional)

Keys can be configured directly inside the interface:

1. Open the Settings menu on the overlay.
2. Enter your Deepgram key under Deepgram API Key (console.deepgram.com).
3. Enter your Groq key under API Key (Groq/DeepL) (console.groq.com) if utilizing Groq LLM or Groq Whisper.
4. Preferences automatically persist to your local appsettings.json file.

Note

All credentials and configurations remain strictly stored on your local machine within appsettings.json and are never transmitted to external third-party servers.

---

Usage Guide

1. Configure Languages:
   • Set the source language (e.g., Russian, Japanese, English) or select Auto Detect.
   • Set the target output language (e.g., English, Indonesian).
2. Select Audio Source:
   • Choose Default System Playback to capture audio from your system output (games, media players, web browsers).
   • Alternatively, choose a specific microphone input or individual application process.
3. Initiate Translation:
   • Click Start Audio Translate to begin live audio subtitle generation.
   • For on-screen text, click Select Screen Area, outline the target region, and click Start Live OCR.
4. Customize Display:
   • Reposition the subtitle box by clicking and dragging it to your preferred position.
   • Adjust caption font sizing (Small, Medium, Large, Extra Large) in the Settings pane.

---

Architecture and Technologies

• Runtime: Windows Presentation Foundation (WPF) on .NET 10.0 Windows
• Audio Processing: NAudio (WASAPI loopback capture and endpoint enumeration)
• Speech-to-Text Engines:
  • Deepgram WebSocket Nova-2 API (Real-time streaming with interim tokens and diarization)
  • Groq Cloud Whisper Large-v3 API
• Translation Providers:
  • Google Translate GTX Client Endpoint
  • Groq LLM API (Llama 3.3, GPT-OSS, Mixtral)
• Optical Character Recognition: Windows.Media.Ocr (Hardware-accelerated native Windows OCR)
• User Interface: Dark glassmorphic design with native WPF hardware acceleration and non-blocking mouse hit-testing.

---

Contributing

Contributions, issues, and feature proposals are welcome:

1. Fork the repository.
2. Create your feature branch (git checkout -b feature/NewFeature).
3. Commit your changes (git commit -m 'Add NewFeature').
4. Push to your branch (git push origin feature/NewFeature).
5. Open a Pull Request.

---

Privacy and Data Handling

• Local-First Audio & OCR: Audio loopback buffers and desktop screen captures are processed strictly in volatile system memory (RAM) and immediately discarded. No recordings are ever written to disk or sent to developer servers.
• Zero Background Telemetry: The application does not collect background metrics, analytics, or user behavior.
• Voluntary In-App Feedback: When you choose to submit a bug report or suggestion:
  • An anonymous local Device ID (ID-XXXX-XXXX-XXXX) and diagnostic metadata (OS build and app version) are included so you can monitor resolution status (Pending, In Progress, Fixed) directly within the app.
  • Submitting contact email is completely optional for receiving developer replies.
• For complete terms, data security policies, and third-party API disclosures, please consult our full Privacy Policy and Terms of Service (PRIVACY.md).

---

License

This project is licensed under the terms of the MIT License. See the LICENSE file for details.