Local by design
Speech recognition and cleanup run on the Windows PC after explicit model setup. Official model downloads are checked against catalog SHA-256 values.
VoxInputSpeak naturally. Write anywhere.
VoxInput turns speech into clear, polished text in almost any Windows app—processed locally on your PC.
Download for WindowsWindows 11 · 64-bit · MSI · No account requiredDownload didn’t start? Try again.This release is unsigned; Windows may show Unknown publisher.Download, licensing and support details →
VoxInputDictationLocal dictation for Windows
ReadyIllustrative previewFocus a text field, then dictate.
Ctrl + Alt + SpaceLatest result
Send the revised proposal before Friday, and include the updated implementation notes.
Local processingNo paid AI APIWorks offline after model setupZero Retention option
How it works
Cursor. Shortcut. Speak.
A local speech-to-text pipeline makes offline voice typing available after model setup while safety checks stay around the workflow.
- 01
Place the cursor.
Focus an editable field in the Windows application where the text should appear.
- 02
Hold the shortcut and speak.
Use the configurable global push-to-talk shortcut, then release it when you are done.
- 03
Polished text appears.
VoxInput transcribes, conservatively cleans, verifies the target, and inserts the result.
Illustrative text cleanup
Natural speech in. Clear writing out.
This website example does not record or transcribe speech. It shows predetermined text to illustrate punctuation and cleanup; it is not a product speed or accuracy test.
EmailDocumentsBrowser formsCRM notesSupport tools
Raw speech
um send the revised proposal before friday comma and include the updated implementation notes
Clean text
—
Privacy architecture
Choose how your dictation is handled.
Review the product’s processing and retention options. Website analytics and the inquiry form are separate from the desktop dictation workflow.
Speech and cleanup stay on this device.
WASAPI captures audio locally, Whisper transcribes locally, and Qwen or a conservative deterministic fallback cleans and formats the result locally.
Security posture
Controls you can inspect. Claims you can verify.
We use NIST Cybersecurity Framework 2.0 to organize product security work. It is a guide for ongoing improvements, not a certification or independent audit.
Retention choices
Local Only can keep transcript history. Zero Retention avoids saving new session transcripts. Existing history and text inserted into other apps need separate deletion and handling.
Release transparency
The current public MSI is unsigned. Its version, size and SHA-256 are published below; code signing and stronger release evidence remain priorities.
Local history is stored in SQLite and is not encrypted by the app. Protect the Windows device and account when dictating sensitive information. Read the security scope and current gaps →
System requirements
For Windows 11 on a 64-bit Intel or AMD PC.
- Windows 11, 64-bit
- Administrator approval for the per-machine MSI
- Internet access may be needed for explicit first-run model downloads
- Compatible microphone and editable Windows target
Release manifest
- Version
- 0.1.5
- Installer
- MSI
- Architecture
- x64
- File size
- 237 MB
- Public release
- 2026-10-09
- Signature
- Not signed—Windows may show Unknown publisher
SHA-256
0EC30E775BFF6A4F6F9381EB4D5DEE0CDBAAB687685AA26BAA2F09E5654BE9A2Install
One installer. Explicit model setup.
Check your Windows version and installer details, then allow time and disk space for the model setup.
- 1
Download the MSI.
Download the version shown above. Its public release page and SHA-256 let you check the file you received.
- 2
Approve installation.
Windows requests administrator approval because VoxInput installs per machine.
- 3
Choose and install a model.
VoxInput recommends a local model for the PC and verifies explicit downloads with SHA-256.
- 4
Focus a field and dictate.
Use the shortcut in an editable Windows application and keep Copy Last available for recovery.
If an insertion cannot complete safely, use Copy Last and paste manually.
Before you install
Download details and practical help.
Public download and licensing
Version 0.1.5 is publicly downloadable here without website checkout. “No paid AI API” describes a technical dependency; it does not specify a software license.
A product price, trial duration and commercial license are not published here. For business deployment or licensing terms, request written clarification before relying on the app for a rollout.
Ask about licensing or support →On a phone, Mac or unsupported PC?
This installer targets Windows 11 x64. It is not a mobile, macOS or native Windows ARM release. Save the setup page and open it on a compatible Windows PC.
Plan for the first run
The installer size excludes any additional model downloads. Model choice affects memory, storage and processing speed. Minimum RAM, free disk space and model download sizes are not yet published on this page; ask about your hardware before a managed deployment.
Use a short, non-sensitive phrase in an editable field for your first check. Confirm the text before using it in business records.
Installation or dictation trouble?
Check the Windows version, microphone access, selected model and editable target. If insertion fails, try Copy Last and paste manually. To remove the app, use Windows Settings → Apps → Installed apps → VoxInput → Uninstall.
Support requests should include the app version, Windows version, the step that failed and an error description, without audio, transcripts or private records. Support response times and update commitments are not advertised.
Check the release
The current installer is unsigned. A checksum checks file identity; it does not replace a publisher signature or prove that software is safe. Do not disable organizational security controls to install it.
View the published version and release notes ↗FAQ
Straight answers before installation.
Does VoxInput require an account or paid AI API?
No. The core workflow does not require an account or a paid AI API. Speech recognition and cleanup run locally with downloaded models or the deterministic fallback.
Does it work without an internet connection?
Yes, after the required local models are explicitly downloaded and installed. The installer itself does not bundle every model, so first-run setup may require an internet connection.
Where can I dictate?
Place the cursor in an editable field in almost any Windows application. VoxInput verifies that the original field is still editable and safe before insertion.
What happens if text cannot be inserted safely?
VoxInput preserves the latest transcript. You can use Copy Last and paste it manually. Password, read-only, or changed fields are not treated as safe insertion targets.
What do Local Only and Zero Retention mean?
Both keep speech and cleanup on the device. Zero Retention additionally disables transcript, statistics, and failed-audio persistence for the session policy.
Is VoxInput SOC 2 or ISO/IEC 27001 certified?
No. We use NIST Cybersecurity Framework 2.0 as a guide for security improvements, but do not claim a SOC 2 report, ISO/IEC 27001 certification, or an independent product audit. See the Trust Center for current scope and limitations.
Is the current public installer signed?
No. The current public installer is not Authenticode-signed, so Windows may display an Unknown publisher or reputation warning. Verify the published SHA-256 before opening it. A future release should be signed and timestamped with a trusted certificate.
Private AI dictation for WindowsSpeak naturally. Keep control.
Windows 11 · 64-bit · MSI · No account required
Download for Windows