Break Every Language Barrier
Aurora Subtitles is a powerful desktop application that transcribes and translates audio in real-time, displaying subtitles as a customizable overlay on your screen. Powered by OpenAI’s Whisper and Meta’s NLLB models, it processes everything locally on your machine — no cloud, no subscriptions, no data leaving your PC.
Who Is It For?
- Gamers: Understand voice chat in any language while gaming. Subtitles appear as a transparent overlay that doesn’t interfere with gameplay.
- Content Creators: Follow international streams, podcasts, and videos with instant translated subtitles.
- Business Professionals: Navigate multilingual meetings and video calls effortlessly with real-time captions.
- Students & Researchers: Access lectures and educational content in any language with AI-powered accuracy.
- Accessibility: Provide real-time captions for hearing-impaired users in any context.
How It Works
- Install & Select Your Model — Choose from Whisper models ranging from ultra-fast (Tiny, ~43MB) to high-accuracy (Large-V3-Turbo, ~574MB). Models download automatically.
- Configure Your Languages — Select the source and target languages. Aurora Subtitles supports 200+ language pairs.
- Start Subtitling — Click START or use your global hotkey. Subtitles appear instantly as a sleek overlay on your screen.
Powered by Leading AI
- OpenAI Whisper: State-of-the-art speech recognition that understands accents, technical jargon, and natural speech patterns.
- Meta NLLB: No Language Left Behind — a translation model capable of translating between hundreds of language pairs with remarkable accuracy.
- Silero VAD: Intelligent voice activity detection that only processes when someone is speaking, saving resources.
Complete Customization
Tailor every aspect of your subtitle experience:
- Font family, size, color, and weight
- Outline color and thickness for readability
- Background color and opacity
- Window position (top, bottom, or custom coordinates)
- Click-through mode for uninterrupted workflow
Built with Rust and optimized for Windows:
- Latency: 300-800ms from speech to subtitle
- CPU Usage: 5-15% with lightweight models
- GPU: 50-70% faster with NVIDIA CUDA
- Memory: ~500MB with base model
