ARIA

ARIA turns audio playing on a Windows PC into live subtitles for videos, games, browser playback and calls. It's for people who need captions across apps or want translations alongside the original speech.

• Local speech recognition: The Full package uses Faster-Whisper with Whisper Large-v3 for multilingual transcription, including Korean, Spanish and French, with an NVIDIA GPU recommended for this mode.
• Fast caption updates: Realtime mode uses Sherpa-ONNX for Chinese and English and Vosk for Japanese, suited to conversations and streams where delays matter.
• Translation alongside captions: Independently movable and resizable overlays keep original and translated text visible together; the Full package includes NLLB for offline translation, while Google Translate, Bing Translator and Youdao receive recognized text through their online services.
• Audio capture and transcripts: The Full package can transcribe system audio or a microphone, and the app can save transcripts and detailed session logs.

The Full package runs on Windows 10 or 11 and includes models for offline recognition and translation. Windows 11 Live Captions provides another recognition option without NVIDIA hardware or CUDA, with language availability determined by Windows.

The smaller Lite package requires Windows 11 version 22H2 or later and limits recognition to Windows Live Captions, with a translation overlay. ARIA is open source under GPL-3.0.