Whispering is a free, open-source transcription tool that has changed shape. The standalone braden-w/whispering repository was archived on 14 February 2026 and the app now lives inside Epicenter, a local-first platform from the same author where your data sits on your machine as plain Markdown and SQLite. The workflow is unchanged — press record, speak, optionally run the transcript through a transformation, then copy or deliver it — but where you install it and what it's part of are different.
At a glance
Platforms
Browser (whispering.epicenter.so) and the Epicenter desktop app (Tauri: macOS, Windows, Linux)
Pricing
Free. Bring your own API keys for cloud models
Free tier
The whole app
Processing
Local GGUF on desktop; cloud providers or self-hosted endpoints if you prefer
Offline
Yes, on the Epicenter desktop app
Licence
AGPL-3.0 for the shipped apps, MIT for the developer toolkit
Self-hosted transcription endpoints; the wider Epicenter app ecosystem
What it does well
Bring your own everything. Point it at OpenAI, Groq, a self-hosted endpoint, or run local GGUF models on desktop. You're not locked to one vendor's pricing or privacy policy.
Runs in a browser. The hosted version needs no install, which is genuinely useful on a locked-down work machine.
Costs nothing but tokens. No subscription. If you use cloud models you pay the provider directly at cost, which for light use is cents a month against $10–15 for a commercial app.
Sensible licensing. AGPL on the apps, MIT on the toolkit, so you can build commercial products on the libraries without the copyleft.
Local-first by design. Epicenter's whole premise is data on your machine in open formats. The privacy story is architectural, not a setting.
Where it falls short
The migration is a real cost. The archived repo is read-only and links to it are stale. Anyone who bookmarked the old app has to find and install Epicenter instead, and the docs are still catching up.
You install a platform to get an app. The only native runtime is the Epicenter desktop host. If you wanted a small dictation utility, you're now adopting something broader.
Local models are desktop-only. The browser build has to use cloud providers, so the no-install path and the privacy path don't overlap.
BYOK is friction. Getting an API key, funding it, and pasting it in is a real barrier compared to signing up for a $10/month app.
Not a dictate-into-any-app tool. The flow is record then copy or deliver, not hold-a-key-and-type-at-the-cursor. Handy is the closer match if that's what you want.
Pricing
Free. Cloud transcription costs whatever your chosen provider charges, billed by them.
Who it's for
Developers and privacy-minded users who are comfortable managing API keys and want transcription without a subscription or a vendor relationship — especially anyone already interested in Epicenter's local-first, own-your-data premise. If you just want to press a key and have words appear in Slack, Handy is free too and does exactly that.
Last verified: 9 August 2026. Note: the original standalone repository was archived in February 2026; this entry describes the Epicenter successor.
Similar to Whispering
Whispering: Now part of Epicenter — free transcription in the browser or on your desktop – Dictation Database
