Skip to content
Nockerl
NockerlVoice

Speech to text for macOS, through omni models.

Free and open source. The audio goes to an omni model rather than a dictation engine. Bring your own OpenRouter key, or point it at a local endpoint.

DownloadView source

macOS 14 Sonoma or later. Free and open source, MIT licensed.

Install with Homebrew

brew tap dizyx/tapbrew install --cask nockerl-voice

Verify the download

These check the build provenance, the SHA-256 checksums, and Apple's notarization.

gh attestation verify NockerlVoice-*.dmg --repo dizyx/nockerl-voiceshasum -a 256 -c checksums.txtspctl -a -vvv "/Applications/Nockerl Voice.app"

How it works

  1. 01

    Double-tap Right Command

    From any app, over any window. Command shortcuts like copy and paste keep working normally.

  2. 02

    Speak

    A click-through pill shows a level meter and elapsed time. It never takes focus from the app you are in.

  3. 03

    Tap once to stop

    The text comes back punctuated and with filler words removed, and is pasted where the cursor is.

NockerlVoiceNockerl Voice

Dashboard: transcription and word counts, total and longest recording time, most frequent words, and a thirty day chart.

It keeps count

Words, total time, longest recording, and what you say most.

The Styles pane: built-in styles above two the user wrote themselves.

Styles are prompts, and they are yours

Three built in. Write your own, and switch between them from the keyboard mid-recording.

The Vocabulary pane: custom terms with the mishearings each should be corrected from.

Teach it the words it gets wrong

Add a term with the ways it gets misheard, and it comes back spelled your way.

Searchable history: every transcript with the style that produced it and when it was recorded.

Everything is searchable, and stays on your Mac

A failed request leaves a row you can retry, and a paste that does not land leaves the text ready to copy.

Bring your own transcription

OpenRouter

Every audio-capable model in its catalogue, on the provider you pick.

Custom endpoint / Local models

Any OpenAI-compatible endpoint: something on your own machine, or another service you would rather use. Styles and vocabulary need an omni model that follows instructions; a plain transcriber returns raw text.

Transcription settings on the OpenRouter tab: zero data retention on, with a chosen model and provider.Transcription settings on the Custom tab, pointed at a local endpoint, API key marked optional.

Privacy

What stays on your machine