Dictation · checked 2026-10-19

Can I vibecode Yaps?

KINDA· weekend project

Yaps bundles several voice tools that are each a short project today: hold a key and paste the transcript, turn an audio file into text and captions, read selected text aloud, keep notes as Markdown. On one Mac, with open models like Whisper and Kokoro, an agent gets all four working in a weekend. The rest is breadth: keyboards for Android and iPhone, the same app on Windows and Linux, speaker identification, translation, voice cloning and note sync between devices are separate builds.

$15/moprice

$180you'd save a year

weekendbuild time

—people replaced it

Before you start

What you need on hand.

DifficultyIntermediate · Some system setup or keys.

TimeA weekend; macOS permissions and model downloads take the longest

Claude Code, Codex or CursorAny of them can run the prompt.
A Mac with Apple Siliconmlx-whisper runs on Apple Silicon only; Intel Macs need a different Whisper backend.
Python 3.12, uv and ffmpegInstall ffmpeg with Homebrew; the agent needs it for video files.
Microphone and Accessibility permissionsGranted to the terminal you launch from; global hotkeys and pasting fail silently without them.
Disk space for modelsFrom a few hundred MB for a small model to several GB for the large one.

Before you buildYaps Free is phone-only voice typing, notes and read-aloud, and every desktop plan starts with a 7-day trial. If voice typing on a phone is all you need, the free tier already covers it.

1. Install one agent: Claude Code, Codex, Cursor.
2. Make an empty folder and open it in the agent.
3. Paste the prompt below, press enter, and answer its questions as it builds.

The prompt

Build me a local voice toolkit for the Mac to replace Yaps. Requirements:

- macOS 14+ on Apple Silicon, Python 3.12 with `uv`: `uv run yaps` starts a menu bar app (`rumps`) and a small
page at http://127.0.0.1:4871 for notes and settings.
- Dictation: hold a global hotkey (`pynput`, default Right Option, rebindable) to record with `sounddevice` at
16 kHz mono. On release, transcribe with `mlx-whisper` (model picker: small, large-v3-turbo; downloaded on first
use into ~/Library/Application Support/Yaps/models), then paste at the cursor via the clipboard and an
AppleScript Cmd+V and restore the old clipboard. Audio stays in memory.
- Cleanup, off by default: remove filler words and fix punctuation through an OpenAI-compatible endpoint
(LLM_BASE_URL, LLM_MODEL, LLM_API_KEY in ~/.config/yaps/.env; local Ollama works).
- Files: `yaps transcribe <file>` runs any audio or video file (ffmpeg extracts the track) through the same model
and writes a .txt and a word-timed .srt next to it. No speaker labels.
- Read aloud: select text anywhere, press a second hotkey, and `kokoro-onnx` speaks it in the chosen voice;
`yaps speak <file.md> --out out.wav` renders a document to WAV.
- Notes: each dictation can be saved as Markdown in ~/Yaps/notes/YYYY-MM-DD-HHMM.md. The page lists, edits and
searches them (SQLite FTS5 in ~/Yaps/index.db); the plain files stay the source of truth.
- Single user. No accounts, no telemetry, local-first. Network only for model downloads and the optional LLM endpoint.
- Out of scope: Windows, Linux, iOS and Android keyboards, voice cloning, translation, speaker identification,
note sync between devices, chat with notes.
- README: granting Microphone and Accessibility to the terminal or the built app, model sizes on disk, and a
30-second test for each of the four commands.

Opens your agent with the prompt filled in, you press enter. Or copy it and paste it into an empty folder.

Share on X

Max MiskowSponsoredAI agents, automations and MVPs, built for your company.Visit site ›

Then keep going

Next prompts, same folder.

The first prompt builds the core. Run these one at a time in the same folder to grow it into your own Yaps.

Extend the dictation toolkit in this folder. Requirements:

- Add modes in ~/.config/yaps/modes.toml: name, system prompt, and a list of bundle ids it applies to (for example com.tinyspeck.slackmacgap).
- Read the frontmost app's bundle id with NSWorkspace when the hotkey is pressed and pick its mode; fall back to Plain.
- Show the active mode in the menu bar while recording.
- Add three example modes: Message, Email and Notes.
- Update the README.

Add custom vocabulary to the dictation toolkit in this folder. Requirements:

- Read ~/Yaps/vocabulary.txt (one term per line) and pass it to Whisper as the initial prompt, capped at 200 tokens.
- Read ~/Yaps/replacements.txt with "heard => written" lines and apply them after transcription, case-insensitive, whole words only.
- Add a menu item "Add last misheard word" that opens a small dialog to append a replacement.
- Unit tests for the replacement rules.
- Update the README.

Add an audiobook command to the toolkit in this folder. Requirements:

- `yaps audiobook <file.md|file.epub> --out book.m4b` splits the text at headings into chapters and speaks each with `kokoro-onnx`, chunked at sentence boundaries.
- Join the chapters with ffmpeg into one M4B with chapter markers and the title as metadata.
- Show progress and an estimated time left in the terminal; resume if interrupted by caching each chapter's WAV.
- Update the README.

What you lose

The Android and iPhone keyboards with voice typing, and note sync between phone and desktop
Windows and Linux apps
Speaker identification and the Studio editor for captions and word timings
Voice cloning and on-device translation in dozens of languages
Chat with your notes and the Save to Yaps browser extension

Why people still pay

People who want private dictation, transcription and read-aloud on every device they own, without wiring models together, pay for the packaging: model downloads, updates for each operating system, the phone keyboards and a sync that doesn't go through a cloud.

Moatexecution polish

Do the math

Yaps or your own copy?

Over

Your usage

Count my time12h ×$/h

Yaps, 1 yr$180

Your own copy, 1 yr$26running + build session

You keep$155pays for itself in the first month

Estimate · Everything runs on your Mac, so the running cost is zero unless you add a hosted LLM for cleanup. The build is token usage on your agent plan.

Leaving Yaps

Take your data, then cancel.

1 · Take your data out of Yaps

1Notes live in a Markdown vault that other apps can read; the notes view also exports to .md and .txt.
2Studio exports generated speech as WAV and captions as SRT.
3Copy the whole vault folder before you cancel; it is plain files, so there is no conversion step.
4Dictation history and settings have no documented export: note any vocabulary and replacement lists you rely on.

2 · Cancel Yaps

1Sign in to your Yaps account and choose Manage billing.
2Cancel the subscription there; access continues until the end date it shows.
3A desktop trial renews into the paid plan after 7 days unless you cancel first.
4The phone apps keep working on the free plan.

Students with a verified .edu email get 50% off the first year of Pro ($72, or $6 a month billed annually).
Referral rewards give a free month to both sides and can stack for up to 12 months.
Yaps says content is processed on your device, so the account is the only thing to delete remotely.
Account deletion is a separate page on the site and is the permanent option.

Checked 2026-10-19 against Yaps's docs: 1, 2, 3, 4

The escape hatch

Don't feel like building it? These are already free.

HandyOpen sourceMIT-licensed desktop dictation that runs offline: press a shortcut, speak and the text is pasted into any field.★ 34k · active Oct 2026 · one-click install
macOS, Windows, LinuxVisit ↗OpenWhisprOpen sourceMIT-licensed voice-to-text with AI cleanup, meeting transcription and notes on macOS, Windows and Linux.★ 9.6k · active Oct 2026 · one-click install
macOS, Windows, LinuxVisit ↗BuzzOpen sourceMIT-licensed offline transcription and translation of audio and video files with Whisper, plus live microphone mode and speaker identification.★ 22k · active Oct 2026 · one-click install
macOS, Windows, LinuxVisit ↗

All 5 free alternatives to Yaps ›

The numbers

Yaps pricing

Plan	Monthly	Annual, per month	What you get
Free	$0	$0	Mobile only: voice typing in apps on Android, iPhone and iPad, voice notes with local search, read aloud with 4 mobile voices.
Pro	$15	$12	Desktop apps for Windows, macOS and Linux: unlimited dictation, read-aloud and transcription of audio and video files, SRT captions, speaker identification on Windows and macOS, chat with your notes, 18 on-device voices, the Studio editor and vault sync.
Enterprise	custom	—	Priority support, volume licensing, centralized billing and private onboarding for a team.

Free tier: Phones and tablets only; no desktop dictation, transcription or captions.
Billing: Monthly or annual. Desktop starts with a 7-day trial that renews into a paid plan until cancelled; cancelling in the account keeps access to the end date.
Hidden costs: Processing happens on your own hardware, so the models are large downloads and need a reasonably recent machine. Voice cloning runs only on supported Apple Silicon Macs, and some read-aloud engines are still marked preview.
Source: www.yaps.ai

Also on the list

Superwhisper$8.49/mo · DictationYES

Wispr Flow$15/mo · DictationYES

Speechify$29/mo · ReadingYES

Questions

Yaps, briefly

Not answered here? Tell us and we'll add it.

Ask a question ›

Free covers voice typing, voice notes and four read-aloud voices on Android, iPhone and iPad, with no account.

Kinda. The core is a weekend build with the prompt on this page, but real gaps remain: The Android and iPhone keyboards with voice typing, and note sync between phone and desktop, Windows and Linux apps.

About $15 a month on the Pro plan, $180 a year (checked 2026-10-19).

The Android and iPhone keyboards with voice typing, and note sync between phone and desktop; Windows and Linux apps; Speaker identification and the Studio editor for captions and word timings; Voice cloning and on-device translation in dozens of languages; Chat with your notes and the Save to Yaps browser extension. If any of those matter to you, keep paying.