Local subtitles for
your own broadcasts.
Scrybe runs whisper.cpp on your machine and pipes live captions straight into OBS as a browser source. No cloud, no API keys, no per-minute billing. Still in early development — use with caution.
Runs locally
All processing happens on your machine. Your audio never leaves the box — no API keys, no cloud latency, no subscription.
Mic or desktop audio
Select any audio device. Caption your own voice, game audio, a VoD you're reacting to - even virtual audio cables.
OBS browser source
Drop localhost:3030 into OBS and the subtitles appear. Tune align, background and transparency live.
Advanced Settings
Swap models, toggle translate, trim silence, etc. Great out of the box, but allows you to customize your experience.
Four steps from install to on-stream.
Scrybe is a Tauri app — install it like any native binary. First-run grabs your selected model, then it serves the overlay on localhost for OBS to pick up.
- 1 Install and launch Scrybe
- 2 Pick an input and a model large-v3-turbo
- 3 Add
http://localhost:3030/app/v1/overlayto OBS 960 × 240 - 4 Press
Start. Speak. Captions appear. go live
Powered by whisper.cpp.
whisper.cpp is Georgi Gerganov's port of OpenAI's Whisper speech-recognition model to plain C/C++. Same models, same weights — just stripped of the Python stack and compiled down to something small enough to ship inside a desktop app.
Scrybe embeds it directly. Audio goes in, text comes out, nothing round-trips to a server. You pick the model size — tiny for a low-end laptop, large-v3-turbo when you want real accuracy — and tune the flags yourself.
More at github.com/ggml-org/whisper.cpp.