Hold, speak, release
Hold Ctrl + Windows to record and release to transcribe. Choose your own one- or two-key shortcut on Windows.
08 / Voice
Turn a spoken idea into text for your next prompt. Hold your shortcut, speak, then release. Sourceweave transcribes on your computer and copies the words to your clipboard, ready to paste into a terminal, thread, or chat.
WHAT IT CHANGES
Hold Ctrl + Windows to record and release to transcribe. Choose your own one- or two-key shortcut on Windows.
A brief In clipboard confirmation tells you the text is ready. Paste where you want it, edit it, and send when you are ready.
Local Whisper turns speech into English text using your CPU or a compatible NVIDIA GPU. Auto mode falls back to CPU when needed.
Move the floating pill where it suits you or minimize it beside Voice in the sidebar. The waveform and recording shortcut stay active.
WHAT IS BUILT IN
A reactive waveform, processing spinner, and clipboard confirmation show each step without opening a full panel.
Revisit, copy, or delete your latest transcripts. Speaking time, word count, and CPU or GPU labels keep activity easy to check.
Choose a microphone, preview a soft recording cue, and save the shortcut and processor that work for you.
Setup requires uv and an internet connection to download the runtime and model. Audio stays in memory on your device; recent transcripts are saved locally. GPU support is an optional download.
HOW IT WORKS
Open Voice and complete the local setup. Pick your microphone and try a recording cue.
Speak naturally while the waveform responds. Each recording can last up to a minute.
Whisper processes the recording locally. Wait for In clipboard to confirm the copy completed.
Paste into your terminal or thread, review the words, and send your prompt when you choose.
KEEP EXPLORING