AI dictation without another subscription.

Verenu is a free, open-source AI dictation app for Windows and macOS. Hold Ctrl Win and speak naturally, and Verenu cleans up the transcript and types it into any app. Bring your own AI provider key or use a local model, with no Verenu subscription.

The real Verenu home screen. The sidebar works, and sample dictation drops into history live.

No telemetryBring your own keyOpen sourceNo Verenu subscription
Why Verenu exists

Good dictation software should not require another monthly subscription.

Verenu gives you a local-first, open-source alternative that you control, whether you bring your own API key or run a local model.

Works the way you actually talk.

Works in any app

Email, docs, Slack, terminal, code comments. If you can type there, you can dictate there.

Hold to talk

One hotkey, hold-to-record. No window to switch to, no button to click. Release and it's done.

Smart cleanup

Off to Strong. Four cleanup levels let you choose how much polishing the LLM does.

See it in action

Your key, your provider

Groq, OpenAI, Google, or AssemblyAI for transcription. Pick cleanup separately.

Compare providers

From hotkey to typed text.

Four steps, every time. The audio and the cleanup pass go straight to the providers you picked, never to a Verenu server.

1

Hold the hotkey

Verenu records locally the moment you press Ctrl Win. Nothing leaves your machine yet.

2

Release to send

Letting go sends that audio straight to the transcription provider you chose, whether that is Groq, OpenAI, Google, or AssemblyAI.

3

A cleanup pass

The raw transcript runs through your chosen model. Filler words and punctuation get cleaned up, and the vocabulary, snippets, and tone from the active context group get applied.

4

Typed where you were

The cleaned text pastes into whatever app had focus. The exchange is saved to your local history.

Choose how much polish.

Same sentence, four cleanup levels. Keep every word, or let Strong cut straight to the point.

Off

"so um i was thinking like we should probably leave a bit earlier you know cause there's gonna be traffic i think"

Light

"i was thinking we should probably leave a bit earlier, cause there's gonna be traffic i think"

Medium

"I think we should leave a bit earlier, there's going to be traffic."

Strong

"Leave early. Traffic."

It gets more yours over time.

Context groups tie the vocabulary, snippets, and tone you want to the apps and sites you use them in. Everything stays stored locally, and switching happens on its own.

Context groups

Knows where you are

Group the apps and sites you dictate in, like work, coding, or messages, and each group carries its own vocabulary, snippets, and tone. Verenu switches automatically.

you say"pushing the branch now"
it typesPushing the branch now.
Vocabulary

Learns your words

Names, jargon, product terms. Correct a mis-hearing once and Verenu remembers. It even auto-learns from the edits you make.

you say"grok's free tier"
it typesGroq's free tier
Snippets

Expands what you repeat

Say a short trigger and the full text appears, like addresses, sign-offs, or the standup update you give every morning.

you say"eta"
it typeson my way, be there in about 10 minutes
Style

Writes like you do

Casual, formal, or all-lowercase-no-punctuation. Pick a style per situation and the cleanup pass matches your tone.

you say"are you free for lunch tomorrow"
it typeshey free for lunch tomorrow? 12 works for me

Bring the provider you trust.

Pick separately for transcription and cleanup. Verenu doesn't lock you into one stack, so you can judge each provider on its own policy.

ProviderTranscriptionCleanupLatency
GroqFree tier available
whisper-large-v3-turbollama-3.3-70b-versatile~1.0s
OpenAIUsage-based
gpt-4o-transcribegpt-4o-mini~1.0s
GoogleFree tier available
gemini-3.5-flashgemini-3.5-flash~3-5s
AssemblyAITranscription only
AssemblyAI APINot availableVaries

AssemblyAI handles transcription only. Use Groq, OpenAI, or Google for the optional cleanup pass. Current release is direct-to-provider; fully local transcription is in development.

What stays. What leaves.

In the current release, your data either stays on your device, or goes directly to the providers you chose, not through a Verenu intermediary.

Stays on your device

  • API keys, in Credential Manager or Keychain
  • Settings, in local app storage
  • Transcription history, in local SQLite
  • Context groups, vocabulary & snippets, with auto-learn data
  • Logs, unless you choose to export them

Leaves your device, only by your choice

  • Recorded audio -> your transcription provider
  • Raw transcript -> your cleanup provider
  • Active context group's vocabulary, snippets & cleanup settings, with cleanup requests
  • Update checks -> Verenu release metadata cache

Once it reaches a provider's API, that provider's policy applies. The full breakdown lives in the README on GitHub.

Stop typing what you could just say.

Free, open source, and running in a few minutes. No card, no sign-up.

Windows
  • Windows 10 & 11, WebView2, not Electron
  • Hold Ctrl Win
  • Keys stored in Credential Manager
Download for Windows
macOS
  • Apple Silicon & Intel
  • Hold Option Space
  • Keys stored in Keychain, needs Mic, Accessibility & Input Monitoring permissions
Download for macOS

Or build from source on GitHub

Verenu is free and open source. Some AI providers may charge for API usage depending on the provider and model you choose.