LiveAudio is deployed and ready to try. v1.1.0 is published on GitHub Releases, with installers for Windows and Linux.

LiveAudio is real-time automatic speech recognition (ASR) for content creators: it captures audio from your microphone or system, transcribes it locally with Whisper, and sends subtitles to OBS Studio over WebSocket. 100% local processing — nothing is sent to the cloud.

What’s inside

Real-time transcription with multiple Whisper sizes (tiny, base, small, turbo).
Voice activity detection to drop silence.
Flexible capture from a physical mic or system loopback.
OBS Studio integration via WebSocket.
Hallucination filtering with customizable blacklists.
Sessions saved as .jsonl and subtitles as .vtt.
Hot-swap device and model switching, no restart needed.

Built in Python with faster-whisper and PyTorch (CPU or CUDA), under the MIT license.

Try it

Grab the installer from GitHub Releases (.exe for Windows, .tar.gz for Linux), or check out the code at github.com/plynte-labs/LiveAudio.

LiveAudio is ready to try

What’s inside

Try it