Edit podcasts and video by editing the transcript, entirely on your device
Emerged finds new products worth a look, daily — what’s new today
browser-whisper
An npm library for local speech-to-text on web pages, no server or API key
Is this yours? Claim itIf you want dictation or captions in a web app without paying per minute or shipping users' recordings to a backend, this gets you there in about three lines.
browser-whisper wraps Whisper, Moonshine and Parakeet speech models behind one small JavaScript API. Pass it an audio or video file, or raw microphone samples, and it yields timed text segments as they are decoded. Nothing leaves the device. Inference uses the GPU where the browser allows and falls back to WebAssembly. Model weights are cached after the first download, and timestamped variants give word-level sync. MIT licensed, by Tanpreet Singh Jolly, with a live demo to try models first.
More like this
A local side process that gives your app a voice and ears via JSON lines
Kokoro's voices with a 7 MB decoder in place of the original 213 MB one
One open 9B audio model for speech recognition, understanding, TTS and editing
Hold a key, speak, and the words land wherever the cursor is
Fast, low-cost speech APIs built on Qwen3 voice models