NaturalDictation is in active Windows development. Early-access list is open →
In active development · On Windows

Speak naturally.
Write anywhere.

Private offline dictation for Windows, built for long-form professional work, local ownership, and dependable daily use.

Join Early Access See What Works Today

Your transcripts, vocabulary, profiles, and model files remain under your control.

Powered by VoxForge

Why NaturalDictation exists

Professional speech deserves software that respects it.

Your words should remain yours

Professional speech may contain private, sensitive, or unpublished material. NaturalDictation is being built so recognition and transcript storage can remain local.

Professional software should remain usable

NaturalDictation is planned around a free Community Edition and a perpetual Pro license rather than required recurring access payments.

Voice notes are not professional dictation

NaturalDictation is intended for sustained writing, correction, vocabulary, commands, and application integration.

Modern users need a modern Windows tool

The product is being developed as a local desktop application with replaceable recognition and cleanup components.

Product demonstration

One utterance, three stages.

Stage 1

Recognized

Raw engine output · lowercase, unpunctuated

Stage 2

Cleaned

Conservative cleanup · brass marks added punctuation

Stage 3

Inserted

Document — draft.docx

Final text in the active application

Cleanup should improve punctuation and grammar without inventing missing meaning.

NaturalDictation cleanup principle

How it works

From microphone to document, entirely on your computer.

  1. Stage 01

    Microphone

    Audio enters through the selected Windows input device.

  2. Stage 02

    Local Recognition

    Speech is processed on the user's computer through a replaceable recognition engine.

  3. Stage 03

    Conservative Cleanup

    Deterministic rules and optional cleanup providers improve formatting while preserving the speaker’s meaning.

  4. Stage 04

    Insert or Save

    Final text can be inserted into the active application and stored in searchable local transcript history.

Current verification

Verified Today

Current verification includes a complete fake-engine runtime path and active real speech-engine integration. Recognition quality is still being measured and improved.

No performance figures are published until they can be honestly measured.

  • Windows desktop application
  • CPU-first architecture
  • Microphone capture pipeline
  • Bounded audio processing
  • Local transcript persistence
  • Searchable transcript history
  • Repeated start and stop handling
  • Stop-time utterance flushing
  • Restart persistence
  • Clean runtime shutdown
  • Replaceable speech-engine boundary
  • Replaceable cleanup-provider boundary

In Development

In Development
  • Production recognition refinement
  • Heuristic grammar cleanup
  • Custom vocabulary
  • Correction workflow
  • Dependable insertion across Windows applications
  • Real-speech regression testing
  • Accessibility refinements

Planned

Planned

Direction, not a schedule — no release dates are being promised.

  • Voice commands
  • Automation
  • Custom profiles
  • Application-specific behavior
  • Domain vocabulary packs
  • Transcript export tools
  • Broader accessibility workflows
  • Alternate recognition engines

Privacy & ownership

Your voice should not become someone else’s data.

  • Offline operation after installation
  • Local transcript storage
  • Local vocabulary and correction data
  • Local model assets
  • No required cloud speech recognition
  • User-controlled exports
  • User-controlled deletion
  • No required subscription for an owned version to continue working

Claims of legal certification, regulatory compliance, or specific encryption features are not made here — only what the architecture is intended to do.

Where each kind of data lives

Stored in a local database on your computer. Searchable inside the application, exportable and deletable by you.

Planned custom profiles will live locally with the rest of your application data, under your control.

Custom vocabulary and correction data are intended to remain on your machine — never uploaded for recognition.

Recognition model assets are downloaded once during installation and run locally. No audio is sent to a server for recognition.

Diagnostic logs stay on your computer and can be reviewed or deleted from within the application.

Exports are created only when you ask for them, saved where you choose, and never transmitted automatically.

Built for sustained work

Made for people who dictate for hours, not seconds.

Writing & authorship

Drafts, chapters, and long-form prose dictated at a natural speaking pace.

Academic research notes

Capture observations and literature notes while reading, without leaving the source.

Sermons & teaching preparation

Develop talks and lessons aloud, then refine the transcript into a manuscript.

Correspondence

Dictate letters and replies directly into the applications you already use.

Administrative documentation

Reports, summaries, and records produced by speaking rather than typing.

Long-form professional dictation

Sustained daily sessions with correction, vocabulary, and local history.

General-purpose tooling. Specialized medical or legal vocabulary is not currently available.

Product editions

Own your dictation software. Rent nothing.

Community Edition

Free

For everyday personal dictation.

  • Core offline dictation
  • Local transcript ownership
  • Essential recognition and writing workflow
Join Early Access

Pro Edition

$149planned perpetual license

For sustained professional work.

  • Advanced professional features
  • Free minor updates within the purchased major version
  • Optional major-version upgrades: $25
  • Existing purchased versions continue working permanently
Join Early Access

Final feature allocation may change before release. No subscriptions, no free-trial countdowns — a purchased version is yours.

Development status

An honest picture of where the work stands.

No progress percentages, no invented milestones — only what has been verified, what is being built, and what comes later.

Verified
  • · Desktop application launches
  • · Microphone pipeline works
  • · Transcript events work
  • · Local persistence works
  • · Transcript search works
  • · Stop flush works
  • · Restart persistence works
  • · Repeated start and stop works
  • · Shutdown works
Current focus
  • · Recognition accuracy
  • · Heuristic cleanup
  • · Real-speech regression corpus
  • · Correction workflow
  • · Reliable text insertion
Later
  • · Voice commands
  • · Automation
  • · Custom profiles
  • · Broader application integration

Questions

Frequently asked questions

NaturalDictation is being built for complete offline operation after installation. Speech recognition, transcript storage, vocabulary, and model assets are intended to remain local.

NaturalDictation is currently in active Windows development. Early-access information will be shared with people who join the list.

NaturalDictation is being developed for CPU-first use on modern Windows computers. Final minimum and recommended specifications will be published after broader testing.

No required subscription is planned. The commercial direction includes a free Community Edition and a $149 perpetual Pro license, with optional paid major-version upgrades.

Answer unpublished until the final licensing structure has been decided.

NaturalDictation is being built as a modern professional alternative for people who need offline Windows dictation, local ownership, correction, commands, and application integration. Full feature parity is a long-term product goal rather than a current claim.

Early access

Be among the first to use NaturalDictation.

Early-access participants receive development updates and testing invitations as NaturalDictation becomes ready for broader use.

  • Development and early-access information only — no marketing lists.
  • Your submission is stored privately and never displayed publicly.
  • Leave the list at any time.

Powered by VoxForge

Please enter your name.

Please enter a valid email address.

Please choose the closest use case.

Consent is required to join the list.

Submissions are stored privately · never displayed publicly

Professional dictation should belong to the person speaking.

Follow NaturalDictation as it develops into a private, offline Windows dictation platform for serious daily work.

Join Early Access View Development Status

Powered by VoxForge