# LiveAudio

> LiveAudio is a free, open-source (MIT) desktop app that generates real-time Whisper speech captions 100% locally and streams them to OBS over a local WebSocket (ws://127.0.0.1:8765). It runs on Windows and Linux. No cloud, no API key, no per-minute cost. Current release: v1.2.0.

Key facts:
- 100% local processing; no telemetry. Internet is needed only on first run to download Python, dependencies, and models — fully offline afterward.
- Transcribes with Whisper (tiny, base, small, turbo); Silero VAD trims silence (configurable onset pre-roll + VAD threshold).
- Capture: physical microphone or system audio (WASAPI loopback is Windows-only; Linux captures the mic).
- Broadcasts subtitle JSON over a local WebSocket to OBS Browser Source or any HTML/WS client on localhost; connections accepted from localhost only, no auth.
- OBS overlay (new in v1.2.0): adaptive vertical "ribbon" subtitle buffer, improved subtitle legibility, and capped reveal-animation timing.
- Free & open-source (MIT). No subscription, no API key. You only pay your own electricity — hardware not included.
- GPU optional; CPU works. NVIDIA CUDA optional but recommended (driver ≥ 525, VRAM ≥ 4 GiB), auto-detected.
- Latency: low and tunable — well under a second on a typical setup.

## Pages

- [Home](https://liveaudio.plynte.com/): What LiveAudio is, who it's for, and the conversion funnel — real-time local captions for OBS on Windows and Linux.
- [Download](https://liveaudio.plynte.com/download/): Real v1.2.0 assets (Windows installer .exe, Linux x64 tarball, pip wheel, source, checksums), OS detection, SmartScreen and checksum notes, and system requirements.
- [Getting started](https://liveaudio.plynte.com/getting-started/): Activation walkthrough — graphical installer vs. uv for developers, first-run download, GPU auto-detect, and profiles.
- [OBS setup](https://liveaudio.plynte.com/obs-setup/): Step-by-step OBS Browser Source guide — add the subtitulos_obs.html overlay at 1920×200, disable "Shut down source when not visible", and set the WebSocket port.

## Docs

- [Docs: Overview](https://liveaudio.plynte.com/docs/overview/): What LiveAudio does and how the local pipeline (capture → VAD → Whisper → WebSocket broadcast) fits together.
- [Docs: Getting started](https://liveaudio.plynte.com/docs/getting_started/): Install, first run, GPU auto-detect, and choosing a profile (Fast / Balanced / Quality / Stable Streaming).
- [Docs: WebSocket & OBS](https://liveaudio.plynte.com/docs/websocket_obs/): The ws://127.0.0.1:8765 broadcast, subtitle JSON payload fields, OBS Browser Source wiring, and custom ports.
- [Docs: Troubleshooting](https://liveaudio.plynte.com/docs/troubleshooting/): Common fixes — SmartScreen, libportaudio2 on Linux, audio device selection, and reconnection.
- [Docs: Config](https://liveaudio.plynte.com/docs/config/): config.json reference — ws_port, profiles, hallucination blacklist, and session output files.

## Reference

- [FAQ](https://liveaudio.plynte.com/faq/): Short, citable answers — is it really local, does it work on Linux, is it free, do I need a GPU, do I need internet, what's the latency, how does it differ from cloud APIs.
- [Comparison](https://liveaudio.plynte.com/vs/): LiveAudio vs. OBS caption plugins, browser-based caption tools, and cloud ASR APIs — local cost, privacy, and latency tradeoffs.

## Resources

- [GitHub](https://github.com/plynte-labs/LiveAudio): Source code, issues, and releases (MIT license).
- [Latest release](https://github.com/plynte-labs/LiveAudio/releases/latest): Always-current download assets and changelog.
- [llms-full.txt](https://liveaudio.plynte.com/llms-full.txt): Expanded, concatenated reference of LiveAudio facts and docs for deeper agent context.