keptwords Coming soon
COMING SOON FOR MAC

Stay in the moment.
Keep the details.

A quieter way to record, transcribe, and revisit your meetings.
Powerful open speech models. A private workspace on your Mac.

Coming soon for MacExplore the workspace

Apple Silicon · macOS 15 or laterOn-device by design

The KeptWords app: a timestamped transcript with editable speaker names, personal notes, and audio playback. The meeting content is labelled demo data.
The real workspace, with sample conversation content. Your library starts with you.
No cloud transcriptionNo meeting botsBuilt for Apple Silicon

A LITTLE LESS NOTE-TAKING

Stay with the conversation.
Find the details later.

From the first hello to the next good idea,
keep the moments that matter in one thoughtful place.

01

Bring the whole room.

Record your microphone, audio from your Mac, or both. Import an existing recording when the conversation has already happened. Save audio alone, or transcribe whenever you are ready.

MICROPHONE + SYSTEM AUDIO
02

Follow every voice.

Get a timestamped transcript with anonymous speaker labels. Add names yourself, keep familiar spellings in your names library, and jump back to the moment you need.

LOCAL ASR + SPEAKER DIARIZATION
03

Make it yours.

Save your thoughts beside the transcript. Search your library, replay the audio, and export the conversation in the format you need.

AUDIO · TEXT · MARKDOWN · JSON · SRT

A MOMENT BEFORE THE MEETING

Sound check.
Then settle in.

Check your microphone and system audio separately. Record a short disposable clip, play each source back, and see exactly which input is receiving samples.

  • Independent source meters during capture
  • Short audio tests with source-by-source playback
  • Explicit test tone and disposable test clips

A moving meter is a useful signal. Listening back is how you check clarity.

KeptWords Audio check with separate microphone and system meters, source selection, and a Start audio check button. Both meters are idle in this sample workspace.
Audio check in the app. Sample workspace; no active capture.

GOOD LISTENERS. GREAT MODELS.

Open models.
A private studio.

Choose the speech model that suits your work. Prepare it once, then transcribe locally with your Mac’s Neural Engine or GPU.

Parakeet TDT v3

Multilingual speech recognition for everyday conversations.

CoreML NEURAL ENGINE

Nemotron 3.5 ASR

NVIDIA’s multilingual speech model, running natively on your Mac.

Metal APPLE GPU

Nemotron Speech English

A dedicated option for English speech recognition.

Metal APPLE GPU

Nemotron 3 Diarization SPEAKERS

Separate up to eight anonymous speakers, locally.

Metal APPLE GPU

Downloads need an internet connection. Transcription stays on your Mac.
Model accuracy varies with language, recording quality, and overlapping speech.

Models are prepared explicitly in the app.

YOURS. ALWAYS.

What’s said here
stays here.

Your recordings, transcripts, and notes live on your Mac. Speech and speaker models run locally. No bot joins your call. No transcription service receives your meeting audio.

NO CLOUD TRANSCRIPTIONNO TELEMETRYYOUR LOCAL LIBRARY

A FEW GOOD QUESTIONS

Before you
press record.

What kind of Mac do I need?+

KeptWords requires an Apple Silicon Mac running macOS 15 or later. Model weights are downloaded separately and need storage and memory; the app shows preparation status and checks memory before processing.

Does KeptWords join my meetings?+

No. KeptWords records audio through native macOS capture. Choose your microphone, audio playing on your Mac, or both. Let everyone in the conversation know before you record.

Does it work offline?+

After the models you need have been downloaded and prepared, transcription runs locally. Internet is needed for initial model downloads and to check or download app releases.

Does it know who’s speaking?+

Diarization groups speech into anonymous speaker labels; it does not identify a person. You can name speakers yourself. Overlap and noisy recordings can affect results, and live-preview labels are provisional.

Can I bring my existing recordings?+

Yes. Import audio into your local library; KeptWords copies it so your original stays untouched. You can export transcripts as plain text, Markdown, JSON, or SRT subtitles, and save your original audio independently of transcription.

When can I try KeptWords?+

Coming soon. We’re preparing KeptWords for its first public Mac release. Downloads and the GitHub repository aren’t available publicly yet. There’s no announced release date; this page will carry the download when it’s ready.

COMING SOON FOR MAC

Make room for
the conversation.

Your Mac. Your meetings. Your words, well kept.

Coming soon for Mac GitHub · coming soon

We’re getting KeptWords ready for you. Public downloads and GitHub access will be announced here.

Apple Silicon · macOS 15 or later