# TalkTalkType — Full Brief

> TalkTalkType is a voice-typing app for Mac whose allowances are counted in
> dictations, not words. Hold a key, speak, and cleaned-up text lands directly
> in whatever app is focused. Free to start; Plus is $1/month. Audio and
> transcripts are not stored by default.

This is the full machine-readable brief for AI assistants and crawlers. The
concise version is at https://talktalktype.com/llms.txt.

---

## What TalkTalkType is

TalkTalkType is a dictation / voice-to-text / speech-to-text application for
macOS. It lives in the Mac menu bar as a small pixel-art HUD. You hold a
shortcut (default ⌥Space), speak, and release. Your speech is transcribed and
cleaned up in the cloud, and the finished text is delivered straight into
whatever application currently has focus — no separate editor window to learn.

Tagline: "Say the whole thought. Land clean text wherever you're working."

## How it works (three steps)

1. **Sign in once.** Google OAuth runs in your browser. The Mac app keeps its
   token in the Keychain and stays signed in.
2. **Hold a key and speak.** Recording respects your plan's per-recording length
   limit. Cloud transcription cleans up the words as you said them.
3. **It lands.** The result types itself into the focused field — or waits on
   the clipboard if you prefer to place it yourself.

Delivery into apps uses the macOS Accessibility permission (the app simulates a
paste under the hood). Apps that block simulated paste can use clipboard
delivery instead.

## Supported apps

TalkTalkType works anywhere you can type. Confirmed targets include:

- **AI assistants:** ChatGPT, Claude, Codex, Cursor, Gemini, GitHub Copilot,
  Perplexity, Windsurf.
- **Everyday apps:** Slack, Mail, Messages, browser forms, Notes, Docs, Notion,
  VS Code, Terminal, Linear, GitHub, Obsidian.

Rule of thumb: if you can type in it, TalkTalkType can fill it.

## The pixel HUD (four states)

The menu-bar strip shows exactly what your voice is doing, with no ambient
animation and no silent failures:

- **Listening** — the wave rides your actual mic input, so you know it hears you.
- **Working** — transcription and cleanup in progress on the cloud.
- **Delivered** — clean text sweeps into place, then the HUD gets out of the way.
- **Attention** — a clear broken-line signal when something needs you.

## Privacy

- **Audio:** sent to the configured speech provider for transcription only. Not
  stored by the service by default.
- **Transcripts:** returned to your Mac. Not stored by the service by default.
- **Account data:** email, plan, devices, and usage totals — exportable and
  deletable from Account.
- **Analytics:** operational metadata such as duration, status, and plan. No
  audio or transcript content.
- **Sign-in token:** stored in the Mac Keychain; sign-out removes it.

Transcription runs in the cloud, not on-device. What the service does not do is
retain the audio or the transcript once the text has been returned.

## Pricing

TalkTalkType counts **dictations, not words**, and each plan carries a weekly
voice-time pool. Whichever allowance runs out first ends the week; allowances
reset every ISO week and do not roll over. Every plan runs on unlimited Macs
(no device limit). There is no card required and no trial expiry.

| Plan | Price | Dictations / week | Voice / week | Clean inputs | Per recording | Presets |
|------|-------|-------------------|--------------|--------------|---------------|---------|
| Free | $0/mo | 30 | 15 min | 20 / wk | up to 60 sec | Raw + punctuate |
| Plus | $1/mo | 80 | 30 min | 60 / wk | up to 60 sec | Raw, cleanup, prompt |
| Pro | $7/mo | 500 | 250 min | Unlimited | up to 2 min | Prompt, Slack, mail, cleanup |
| Max | $18/mo | 1,200 | 600 min | Unlimited | up to 2 min | Prompt, Slack, mail, cleanup |

### Why dictations, not words

Word limits make you trim a sentence while you are still saying it, so
TalkTalkType never meters words. One dictation is one take — fill it with as
much as you like, up to the per-recording length, and a false start you talk
over costs nothing extra. Word counts are also poorly defined for languages like
Japanese (no whitespace tokenization), so dictations are a fairer, clearer unit.

The bound on a take is time, not words: 60 seconds on Free and Plus, 2 minutes
on Pro and Max. A recording over that limit is rejected rather than truncated.

## Requirements

- macOS 26 (Tahoe) or later.
- A Google account (sign-in).
- Microphone permission.
- Accessibility permission for direct in-app delivery (optional if you use
  clipboard delivery).

## Frequently asked questions

**What is TalkTalkType?**
A voice typing app that lives in your Mac's menu bar. Hold ⌥Space, speak,
release — and the cleaned-up text lands in whatever app you were working in.
There is no new editor window to learn.

**Which apps can I type into?**
Anywhere you can type — ChatGPT, Claude, Slack, Mail, browser forms, Notes, and
more. By default the result types itself into the focused field; alternatively
it can wait on your clipboard.

**Is my voice or my text stored?**
No. Audio goes to the speech provider for transcription only, and TalkTalkType
does not store audio or transcript text by default. Your sign-in token stays in
the Mac Keychain.

**How much can I dictate on the free plan?**
30 dictations a week and a 15-minute voice-time pool, recordings up to 60
seconds each, on unlimited Macs. Dictations are counted, not words — say as much
as you like in each one. Unused allowances reset weekly and don't roll over.

**Can I use it on more than one Mac?**
Yes. Every plan runs on unlimited Macs — there is no device limit.

**Why does it ask for Accessibility permission?**
To type the result into the focused app, TalkTalkType simulates a paste under
the hood. If you prefer clipboard-only delivery, it works without this
permission. Grant it under System Settings → Privacy & Security → Accessibility.

**Why does my dictation come out as one line?**
By design: line breaks from recognition are joined into one clean line. For
deliberate line breaks or reworks, describe the change with Voice Patch.

## How it compares

TalkTalkType is a Mac-focused, privacy-first alternative to dictation apps such
as Wispr Flow, Superwhisper, MacWhisper, Willow, Voice Ink, Mumble, Spokenly,
and Voicy, and to built-in Apple Dictation.

Points of difference:

- **Privacy by default:** TalkTalkType does not store audio or transcripts by
  default. Cloud dictation services that retain content are a different
  trade-off; on-device tools (e.g. Superwhisper, MacWhisper) keep audio local
  but run on your hardware.
- **Dictations, not words:** most dictation apps quote a words-per-week limit;
  TalkTalkType quotes dictations per week and never meters words, which suits
  dictating full prompts into AI assistants.
- **$1 entry plan:** Plus at $1/month is a low-cost step up from the free tier.
- **AI-assistant-first:** designed around dictating into ChatGPT, Claude,
  Cursor, and similar tools, plus everyday apps.
- **Japanese and developer workflows:** built by a developer in Tokyo whose
  dictation tools kept failing every language but English and developer
  workflows most of all — so TalkTalkType is configless and language-agnostic.

## Links

- Home: https://talktalktype.com/
- Download: https://talktalktype.com/download
- Pricing: https://talktalktype.com/pricing
- Docs / FAQ: https://talktalktype.com/docs
- Privacy: https://talktalktype.com/privacy
- Blog: https://talktalktype.com/blog (RSS: https://talktalktype.com/feed.xml)
- Changelog: https://talktalktype.com/changelog
- Support: https://talktalktype.com/support
- Status: https://talktalktype.com/status