LISTEN CLOSELY. KEEP THE SOURCE.
From a recording
to something
worth keeping.
Transcribe. Follow the timestamps. Correct what you hear. Turn the useful parts into notes that always point back to their source.
“A careful reader can follow
the original recording.”
YOUR WORKBENCH
A clear path through the recording.
Opening a private workspace…
1. Choose your recording
English speech · MP3, M4A, WAV, FLAC, OGG, WebM or video with audio · up to 64 MiB / 60 minutes.
No transcription starts during inspection. For YouTube and podcasts, use a recording you may process or the local CLI. Hosted URL downloading is not enabled.
A bounded sample is available once per workspace/IP, subject to global caps. All selected deliverables must complete before paid credit is consumed.
2. Follow the work
Your jobs will appear here.
Credits & private recovery
Keys restore Sonaweft credits and unexpired files. This balance is separate from Pageweft. Do not send keys or recordings to support.
3. REVIEW & TAKE IT WITH YOU
Keep your ears in the loop.
Original machine transcript
Original text stays unchanged. Saved corrections are exported separately. AI suggestions are advisory.
AI-derived notes
Claims to investigate
Optional source lookup sends only the selected claim to Crossref. Matches are unreviewed scholarly metadata, not proof.
Portable instascript-export/v1 bundles include source hashes and segment references for Obsidian or the planned Noteweld importer. That importer is a later feature.
Your knowledge should outlive the tool.
Use files in your own editor, Obsidian vault or language-model workflow. Run the CLI locally when hosted processing does not suit your data. Files expire after 48 hours; export what you want to keep.
Working from PDFs? Open Pageweft ↗