Speak now. Find the thought later.
Record a thought, create a local transcript, search it with the rest of your work and return to the exact moment in the audio.
From recording to searchable context
Debrief: Steering Committee Consensus
Recorded on MacBook Air · Audio: 16kHz mono Opus
Just wrapped up the steering committee debrief with David Kim and Elena Rostova.
Key decision: Aurora PostgreSQL Multi-Region is formally approved as the primary database architecture.
We explicitly rejected the earlier DynamoDB proposal because the compliance audit required strict ACID cross-table consistency.
Target cutover is locked for Q4 with budget capped at €1.2M. Follow-up action item is to review load tests in Frankfurt.
Ask Thread about this voice note:
Four useful stages. One piece of work.
- 01
Speak
Record in a supported browser or attach a supported audio file.
- 02
Transcribe
Whisper resources run locally once they are available in the browser.
- 03
Search
Transcript text joins the searchable material in its Space.
- 04
Ask later
Inspect a cited transcript segment and seek near that moment in the audio.
Your recording can become searchable without a transcription API bill.
Thread uses browser-based Whisper resources for local transcription. Performance and practical recording length depend on the browser, device memory and available model resources.
Read the technical setup and limitsWhat remains useful afterward
- A transcript you can edit
- Timed segments that seek into the recording
- Text that appears in lexical search
- A source that optional Ask can cite
Recording, resources and privacy
Does audio leave my device?
By default, recordings stay in this browser on this device. If you enable sync, permitted recordings can pass through Thread’s encrypted temporary relay to your own devices.
Does transcription need a connection?
The first run needs model resources. Once the browser has retained them, transcription can run locally; browser storage eviction or a failed download can make those resources unavailable again.
Can I jump from text to audio?
Yes. Transcript segments include timestamps. Selecting one seeks playback near that moment in the recording.
Can I correct a transcript?
Yes. Your edited text becomes the current content used by display, search and retrieval while the original machine transcript remains available as historical evidence.