Less typing.
More waffling.

Voice-to-text that works in Slack

Press a hotkey, speak, release — your polished text lands straight in the app you're using.

Download for Desktop iOS — Coming soon

Free & open source · Bring your own API key

👩‍💼
Hey, any updates with the client?
👨‍💻
Fn (Mac)
What you said

Works with everything

ChatGPT
Claude
Slack
Teams
VS Code
Gmail
Notion
Google Docs
Discord
WhatsApp
Cursor
Linear
Perplexity
X
iMessage
Telegram

Voice is not just faster.It's a different league.

Average words captured per minute

Typing
40
Fast typing
70
Waffler
153
3.8× faster than typing

From waffle to structured text in seconds

fn
Step 1

Hold your hotkey

Press your shortcut from any app. The waffle pops up — you're live.

Step 2

Speak naturally

Talk like you normally would. Ramble, pause, waffle — it's all good.

Hello!
Step 3

Release — text appears

Let go and watch your words type themselves out, right at the cursor.

Speaks your language

Add tricky words once. Names, acronyms, internal jargon — Waffler gets them right every time, even when generic speech-to-text models stumble.

My vocabulary
+ Siobhan
+ JSON
+ Postgres
+ macOS
+ |
Names
Shavon Siobhan
Acronyms
Jason JSON
Tech
post grass Postgres
Casing
Mac OS macOS

Set it up in Settings → Vocabulary. One word per line. No regex, no special syntax — just type the word the way you want it written.

See how it works for different tasks

Pick a scenario. Watch one voice memo become the right kind of writing for the job.

You said 0:18 dictation

Hey Sam, um, thanks for sending the report over yesterday, really appreciate it. Quick thought, could we jump on a call tomorrow at 10? Cheers, James.

Waffler wrote

Hi Sam,

Thanks for sending the report over yesterday, really appreciate it.

Quick thought: could we jump on a call tomorrow at 10?

Cheers,

James

Your voice, your business

See exactly where your data goes.

Your voice
Captured locally, never saved
encrypted →encrypted ↓
Your provider
Your API key, your choice
text back →text back ↓
Your clipboard
Paste anywhere

Your audio goes device → provider → clipboard. We're never in the middle.

Frequently asked questions

Everything you need to know about Waffler.

Do I need an account?
No. There is no account needed for Waffler. Download the app, add your API key, and you're ready to go.
Is it really free?
Yes. Waffler itself is 100% free and open source (MIT licensed). The only cost is what your API provider charges per dictation. Pricing below is based on Cerebras Qwen-3 235B (roughly 0.3p per dictation), with the first ~15 dictations per day free via Groq's daily allowance.

What it actually costs per month:

How much you dictateWaffler
(API only)		Wispr Flow
(no BYOK)		SuperWhisper
(sub + API)
Casual (10/day)£0 / month£12 / month£7 + API / month
Moderate (25/day)~£1 / month£12 / month£7 + API / month
Power user (50/day)~£3.50 / month£12 / month£7 + API / month

Wispr Flow doesn't let you bring your own API key, so every word goes through their cloud at a flat fee. SuperWhisper allows bring-your-own-keys only on the Pro plan (£7/month), and the API costs sit on top of that subscription.

Most Waffler users dictate within Groq's daily free tier and pay nothing at all. Even at high volume, the API bill stays a fraction of a subscription, and unlike subscription apps, you only pay for what you actually use. A quiet month costs you nothing (no annual renewal, no auto-charge, no price hikes).
Is my data private?
Waffler has no servers and no backend — it runs entirely on your machine. Audio and transcripts are sent directly from your device to your chosen AI provider (Groq, Cerebras, or OpenAI) using your own API key. All three are US-based and explicitly state they don't train on API data. Waffler never sees or stores your audio. Your transcription history is saved only on your device. See how each provider handles your data →
Which provider should I use — Groq, Cerebras, or OpenAI?
Start with Groq. Its free tier covers most users for everyday dictation and the styling quality on Llama 3.3 70B is excellent. Add a Cerebras key (with $10 of credit) if you want sub-500ms styling on long transcripts or you regularly hit Groq's daily free-tier cap. Add OpenAI only if you want a last-resort safety net for the rare case both Groq and Cerebras are unavailable. Waffler tries them in that order automatically: Groq → Cerebras → OpenAI.
Is the data secure if Cerebras runs an Alibaba model?
Yes. Qwen is open-source (released by Alibaba on Hugging Face for anyone to download and run). Cerebras hosts the weights on their own data centres in Sunnyvale, California — Alibaba never sees the requests. The relationship is the same as Meta open-sourcing Llama and Groq hosting it. All inference happens on US infrastructure under US data protection laws.
Does Waffler update itself?
Yes. From v3.8.5 onwards, Waffler has a Check for Update button in Settings → About. It compares your version with the latest release, downloads the new installer, and relaunches the app automatically. On macOS the update swaps the signed app bundle; on Windows the Inno Setup installer performs a silent upgrade.
Is the app signed?
The macOS build is fully signed with a Developer ID and notarized by Apple — it opens cleanly with no "unidentified developer" warnings. The Windows installer is currently unsigned (code-signing certificates are on the roadmap). If Windows SmartScreen warns on first run, click "More info" then "Run anyway".
How do I get an API key?
It takes about 2 minutes. Start with Groq — it's free and covers most users. Just sign up at console.groq.com, create a key, and paste it into Waffler. Cerebras and OpenAI keys are optional extras for speed or fallback. Full step-by-step for all three: How to Get an API Key.
Mac or Windows?
Both are supported. Head to the download page and pick your platform.

Free. Fast. Private. Yours.

Open source voice-to-text with no account, no subscription, no data collection. Just download and start talking.

Download Waffler

Free & open source · Mac + Windows · No account required