Stream Captions

Stream Captions adds real-time subtitles to live streams, VODs and music — on Kick, Twitch, and any other site you choose to enable. Speech recognition runs entirely on your own device using OpenAI's Whisper model. Your audio never leaves your computer. There is no account, no API key, no subscription, and no per-minute cost. FEATURES • Live captions over the player on Kick and Twitch • Works on any other site — enable it per-site from the popup • Handles live streams, VODs, and audio-only pages like music players • One sentence per line, with a new line when the speaker changes • Non-speech is captioned the way it always has been: [explosion], ♪ Music • Three model sizes: tiny (fastest), base, or small (most accurate) • Choose which corner the captions sit in • Captions are per-tab, so you control each stream independently PRIVATE BY DESIGN No analytics. No telemetry. No tracking. No third-party SDKs. Audio is held in memory only for the few seconds needed to transcribe it, then discarded. Transcripts are never uploaded or saved. The first time you turn captions on, the speech model is downloaded from Hugging Face and cached by your browser. After that it works fully offline. REQUIREMENTS A GPU with WebGPU support gives near-real-time captions on the tiny and base models. Without it, transcription falls back to WASM and is noticeably slower. Expect a few seconds of latency on live streams. Captions need audible audio — a muted player produces silence, though low volume is fine. Free and open source (MIT): https://github.com/sinafayazi/stream-captions