VoiceSnip

Record tab audio, split it by speaker, and export clean clips with matching transcripts.

VoiceSnip — turn the audio in a web page into usable clips Record up to 30 seconds of audio from a browser tab, let it separate the speakers, then export the voice you picked as clean clips — each one with its matching transcript. Three steps 1. On the tab that is playing the video or audio, click the extension icon and press Start recording. 2. Recording stops by itself after at most 30 seconds and processing begins. 3. The panel lists the speakers it found. Preview any of them for 3 seconds, pick the one you want, export. What you get - One WAV per clip, never longer than 10 seconds. The splitter prefers complete sentences of 7–10 seconds and places cut points on pauses. - With a key configured, a matching `.asr.txt` sits beside each WAV, holding the text of that clip. - Files land in the `voice-extract` folder inside your Downloads. You can choose the base file name yourself. Cost: bring your own key Speaker separation and transcription run through your own AssemblyAI account. Signing up is free and no bank card is required. Usage is billed to your account: roughly $0.002 for a 30-second run. The key stays on your machine and never passes through any server of ours. It also works without a key: the extension then splits the recording locally by silence. You get clips, but no speaker list and no transcripts. What this version deliberately does not do - It does not remove background music. Clips keep whatever music and sound effects were in the original. For AI dubbing or speech-synthesis reference audio, prefer dialogue, interviews or podcasts without music. - A single recording is capped at 30 seconds, on purpose: it keeps each run short and cheap. - Overlapping speech is not separated; the seconds where two people talk at once are attributed to one speaker. - DRM-protected subscription streaming sites are not supported. - It does not synthesise speech and does not clone voices — it only prepares the material. Privacy There are no developer servers. Audio is produced only when you start a recording, and with a key configured it is uploaded to your own AssemblyAI account. Your key, settings and state stay on your machine. No accounts, no analytics, no ads. Full details: https://bright103huang.github.io/voicesnip-legal/