Hold, speak, release
Your configured dictation key turns speech into text at the cursor. A quick tap can switch to hands-free mode.
Free · Windows · v1.0.7 · local only
Private voice, right where you type.
Hold a key, speak, release — cleaned-up words land at the cursor. BackFlow runs on your Windows PC. Your audio and the text it produces never leave this machine.
The installer is not Authenticode-signed yet. Windows SmartScreen may warn before launch.
Local · no transfer
Nothing leaves this PC.
BackFlow keeps everything on your computer — the microphone audio, the transcript, the words it types. It does not send that information to us, to a cloud, or to anyone else, at any time. The only thing our servers ever hear from the app is a check for a quality update.
On stage
What it actually does.
Stays on the machine
Speech models run locally. Audio is memory-only by default. There is no account, no telemetry of what you said, and no third-party transcription service.
Quality updates only
The app may ask our servers whether a newer build exists. That check is not a channel for your dictation. An update installs only if you choose to restart and apply it.
Windows 10 and 11
64-bit Windows, a working microphone, and disk space for the local model you pick. NVIDIA hardware helps, and is not required. macOS is next.
How to start
Install, choose a local model, allow the microphone when Windows asks, then use your hotkey. Setup, privacy, and troubleshooting live in Help.
Support
Need a hand? Email support@backstagecoding.com. Share only the diagnostics you are comfortable sending.
Latest cue · 1.0.7
Released 25 August 2026.
Fixed: post-signature artifact verification stays in the signer scope that loads the release validation functions. Full notes in the changelog.