Closed captioning
Early previewClosed captioning is in active development and ships in the Forwarned (Beta) preview ring. Turn it on from the License and Rings page in the desktop app. Its behavior and options can change between releases.
How it works
- The desktop app captures your microphone and streams the audio to a separate caption process that runs a local speech-to-text engine (Whisper). Partial captions update as you speak, then finalize when a phrase closes.
- The caption process is isolated from the app, so if it ever crashes the rest of Corvus Connection keeps running. Finalized captions are saved to a searchable local transcript.
- Captions stay live under load. If the engine falls behind, the oldest audio is dropped to keep latency low, so a few words may be missed during a very busy moment rather than the captions lagging.
Local and private
Captioning runs entirely on your machine. Your audio is transcribed on your PC and does not leave it, apart from the captions you choose to put on screen. There is no cloud service, no per-minute cost, and no extra window to babysit.Choose how it runs
You pick the engine backend that fits your hardware. The choices are Auto, CPU, Vulkan, and CUDA.- CPU is bundled and works offline out of the box. It is the default and the automatic fallback.
- Vulkan is the general GPU path for most graphics cards, and CUDA is the tuned path for NVIDIA. These GPU packs download on demand into a per-user cache and can be removed later to reclaim disk space.
- Auto picks the best backend you have installed. Changing the backend restarts the caption process, and the app shows a short switching state while it does. Captioning is Windows (x64) only.
Using it
- Enroll in the Forwarned (Beta) ring and enable closed captioning.
- Pick your microphone and choose a spoken language (English by default).
- Start captioning. Watch the words in the in-app live caption view, and use the mute hotkey to hold captions when you need to.
