local-first transcription for desktop work
Voice, faithfully written.
Turn spoken work into a written record without inviting a meeting bot or making cloud upload the default. Plainsong captures from the desktop, transcribes locally with Whisper, keeps review close to the source, and exports evidence you can inspect.
Plain by design
Five plain assurances, ruled and kept.
Global hotkey capture i
No attendee bot required ii
Local-first transcripts iii
Review before export iv
Cloud kept optional v
What it holds
Plainsong is not trying to be a calendar guest or a notes portal. It keeps the core loop close: capture speech, transcribe locally, review the record, then export only what should leave.
dictation
Start from a global hotkey, speak naturally, then move clean text into the work you already had open.
The product starts with the fastest path from thought to draft. It should feel more like a quiet input layer than another destination.
local
Run the core speech-to-text path on your machine, with Whisper model management in the desktop app.
Sensitive audio can stay under your control. Hosted services should be opt-in, clearly named, and easy to reason about.
meeting
Record conversations, review transcripts, and keep follow-up decisions tied to the source context.
Use meeting capture for interviews, planning calls, support reviews, and decisions that need a record after the call ends.
review
Clean up who said what with speaker aliases and diarization paths that make transcripts easier to trust.
Review is built around making captured speech searchable, accountable, and ready for handoff without pretending the first pass is final.
export
Package transcripts and verification artifacts for teams that need records they can inspect later.
Export paths should make it clear what is leaving the app, where it is saved, and what source material supports the written record.
assist
Use Ollama-backed review flows for summaries, action items, and project notes without default cloud upload.
Plainsong fits people who want AI help but still care where their audio, transcripts, and follow-up notes live.
Capture loop
Four movements of a single plain line: caught, kept close, shaped, then sent.
Use the hotkey for dictation or begin meeting capture when the conversation needs a record.
Whisper turns audio into text on the desktop path, so private material is not forced through a remote pipeline.
Speaker labels, transcript history, and local AI review turn raw audio into text you can actually use.
Share transcripts and evidence bundles when a decision, commitment, or customer record needs to be checked.
Release ledger
Plainsong is an open-source release track, not a checkout funnel. This page keeps the state plainly visible, a single faithful line of where the work stands, without turning pre-release software into a sales ladder.
Free · open-source · local-first
Plainsong is being prepared as a free open-source desktop app, not a checkout funnel.
Free MIT release track
Built around desktop capture, local Whisper transcription, human review, and deliberate export.
Local by default
Early builds should be treated as evaluation software until release signing and updates settle.
Careful pre-release
Marginalia
Read before trusting a desktop recorder.
Everything stays local and legible. If something below is still unclear, the answer is meant to be plain, write to us and we will annotate further.
No. Plainsong is local-first by design. Core recording, transcription, and review flows are built around local processing.
Yes. Local AI analysis is designed for Ollama-based workflows so your data handling stays under your control.
Yes. Plainsong includes speaker aliasing and diarization pathways to improve transcript readability and verification.
Current launch focus is macOS and Windows. Linux may work in some environments but is not a GA launch target.
No checkout plan is offered through this site. The current app direction is free, open source, and local-first.
Use the contact page for release questions. Public release details should stay tied to the open-source repo and current test builds.
Walkthrough · three movements
The launch media can come later. What matters now is the shape of the work: how a voice becomes a plain, faithful written record you can trust.
Dictate drafts, notes, and follow-ups in place, no browser tab, no upload. The voice is captured on-device, where it stays.
Correct speaker names, scan the commitments, and keep the source context attached. Nothing leaves until you say so.
Send a clean transcript, an evidence bundle, or a local AI summary, a faithful written line, ready to hand off.
Colophon
A free, open-source desk for the spoken word.
Plainsong is being prepared as a local-first desktop recorder: dictation first, meeting capture when you need it, nothing leaving your machine, and no checkout language until there is a real reason to add it.
local-first · no account · open source
