FluentView evidence
Windows voice control

Voice control for Windows, with intent kept visible.

Say the outcome you want. Fluent prepares the Windows route and keeps its progress visible.

Read the short answer
Short answer

Can Fluent control Windows by voice?

Yes, in preview. Fluent can accept a spoken request, transcribe the audio on the Windows device, send the resulting text to a user-configured hosted planner, and run Windows UI Automation tools. The current app exposes progress and cancellation, but it is not yet a finished assistive technology product. The published research and privacy pages document unresolved release and safety gates.

How a spoken request becomes a Windows action

Fluent separates speech recognition from action planning. A constrained Windows listener waits for the Fluent wake phrase. A microphone click or the Ctrl+Alt+, shortcut can also arm one utterance. Chromium captures that command, and a bundled whisper.cpp runtime transcribes it locally.

The text request then enters the same command path as typed input. A hosted planning provider selects from bounded automation tools, such as observing the accessibility tree, opening an app, clicking a named control, or typing text. The app displays action progress and accepts cancellation while work is interruptible.

  • Wake phrase, microphone click, or push-to-talk arms one command.
  • Raw speech audio is transcribed on the device, not by a cloud speech API.
  • The text request is sent to the configured hosted planning provider.
  • The planner uses Fluent tools backed by Windows UI Automation and PowerShell.
  • The overlay shows progress and lets the user cancel an active run.

Natural requests are different from command memorization

Traditional voice control often asks the user to remember an exact command, identify a control by name, or select a numbered overlay. Fluent explores a different layer: the user describes the outcome and the planner decomposes it into smaller semantic actions.

That approach can reduce command syntax, but it also introduces model uncertainty. Fluent therefore treats visibility, cancellation, semantic targets, and consequence-aware approval as core requirements, not optional polish.

Where Fluent may fit

Fluent is being designed for people who want to reduce repeated keyboard and mouse work across changing Windows applications. That may include people navigating pain, fatigue, tremor, injury, limited reach, or simply a workflow where speaking the outcome is easier than relaying every click.

Voice is not the only path. Typed requests use the same surface when speech is private, tiring, unreliable, or unavailable. Optional gaze can add a semantic target when a phrase such as "open that" needs visible context.

  • Cross-app tasks where the desired outcome is easier to say than the exact route.
  • Workflows that benefit from a visible action trace and a nearby cancel path.
  • Users who need voice and text to coexist rather than compete as separate modes.
  • Early testers who can tolerate preview setup and help document failures.

What Fluent does not promise yet

Fluent is not a replacement for every established accessibility tool. It currently requires Windows, a supported hosted planning-provider API key, and developer-oriented setup. The published product record identifies missing first-run provider setup, incomplete installed-app golden paths, gaze recovery gaps, and safety paths that still require hardening.

A production claim must eventually be supported by fresh-install success, repeatable workflow results, consequence fixtures, accessible recovery, signed distribution, and user testing. Until those gates pass, Fluent should be evaluated as a transparent preview.

Direct answers

Questions people ask before trying Fluent.

Does Fluent send speech audio to the cloud?
No cloud speech service receives the raw audio. Fluent transcribes speech locally. The resulting text request is sent to the hosted planning provider configured by the user.
Can Fluent work without a camera?
Yes. Voice and typed requests work without a camera. Gaze is optional and off by default.
Does Fluent work on macOS or Linux?
No. The current automation stack is Windows-specific and uses PowerShell, Win32 behavior, and Windows UI Automation.
Is Windows Voice Access still useful if I try Fluent?
Yes. Windows Voice Access is a mature built-in option for direct voice navigation and dictation. Fluent is a separate preview exploring natural requests, optional gaze context, and planned multi-step work.
Check the record

Sources and product disclosures.

  1. Fluent privacy and safety disclosureCurrent speech, planner, gaze, and action boundaries.
  2. Fluent research methodCurrent evidence status and release gates.
  3. Fluent guide libraryRelated explanations for the current preview.
Written and maintained byJason Matthew Suhari

Creator of Fluent. Product claims are reviewed against the current implementation and published limits.

GitHub profile
Inspect before you trust

Judge Fluent by the evidence.

Read the method, inspect the limits, and evaluate the current preview against published thresholds.

Read the research method