Fn-key push-to-talk
No app switching, no clicking. Use the key already on your keyboard.
Mac dictation · Bring your own key
Voice-to-text without the subscription
ParlaParla turns speech into polished text anywhere on your Mac. Bring your own OpenAI key and pay only for what you use — typically a few cents per hour of dictation.
macOS 15 Sequoia or later · One-time purchase, no subscription
Dictation shouldn't cost $180 a year
Modern dictation apps charge $15 a month or more — for access to the same OpenAI technology you could use yourself. Your voice data flows through their servers. Your workflow depends on their uptime. Your costs never go down.
ParlaParla does it differently. You bring your own OpenAI API key. Audio goes straight from your Mac to OpenAI. Nothing passes through us, because we don't run any servers. You pay OpenAI directly at their standard rates — typically a few cents per hour of dictation.
The result: a Mac-native dictation app that respects your privacy, your wallet, and your workflow.
Three steps to talk instead of type
-
Bring your OpenAI key
Create one free at platform.openai.com. Your key lives only in the macOS Keychain — never on our servers, because we don't have any.
-
Press Fn and speak
Works in any app. Release to insert. Double-tap for hands-free recording. No window to open, no button to click.
-
Text appears, polished
Optional AI cleanup fixes filler words and punctuation in the background. Pick your level — from raw to fully rewritten.
Everything you need, nothing you don't
Works in every app
Slack, Mail, Notion, Terminal, Xcode, Safari — if you can type there, you can dictate there.
Custom dictionary
Names, brands, and technical terms spelled correctly every time.
AI enhancement levels
None, light cleanup, full polish, custom prompt, or translate on the fly.
Multilingual
English-optimized by default, with full support for nine other languages.
Private by design
API key in Keychain. No servers. No telemetry. Sandboxed.
From raw speech to polished text
Choose how much cleanup you want. Here's the same spoken sentence at three levels.
hey joey you still up for coffee i was thinking uh maybe we should leave a bit earlier there might be traffic
Hey Joey, you still up for coffee? I was thinking — maybe we should leave a bit earlier, there might be traffic.
Hey Joey, are we still on for coffee? We should probably leave a little earlier in case of traffic.
Speak your language. Write theirs.
Hold Fn and speak in whichever language is easiest for you. ParlaParla transcribes, polishes, and translates on the fly — the final text is inserted in the language you picked, no tab switching, no copy-paste into Google Translate.
Hi Morten, could we push tomorrow's meeting back by half an hour? Something came up that I need to handle first.
Hej Morten, kunne vi skubbe mødet i morgen en halv time tilbage? Der er dukket noget op, som jeg er nødt til at klare først.
Thanks for taking the time to meet today. I'll send over a revised proposal tomorrow with the updated hours estimate.
Vielen Dank, dass Sie sich heute Zeit für unser Gespräch genommen haben. Ich schicke Ihnen morgen einen überarbeiteten Vorschlag mit einer aktualisierten Stundenschätzung.
Perfect for writing emails, Slack messages, or code comments at native-language speed — in a language that isn't your native one.
We can't see what you dictate. By design.
We designed ParlaParla so that your voice never touches our infrastructure. It can't — we don't have any.
-
No servers
We don't operate any. Your audio goes directly from your Mac to OpenAI.
-
Keychain-only key storage
Your OpenAI API key lives in macOS Keychain and never leaves your Mac.
-
Fully sandboxed
ParlaParla uses Apple's App Sandbox entitlement. We can't read files you don't explicitly grant us access to.
ParlaParla vs. Wispr Flow vs. Apple Dictation
Real numbers. Apple's dictation is free, but stops at the language you set. Wispr Flow is a subscription. ParlaParla charges you what OpenAI charges you — nothing more.
| ParlaParla | Wispr Flow Pro | Apple Dictation | |
|---|---|---|---|
| Subscription | None | $15 / month | Free |
| 1 hour / month | ~$0.30 | $15.00 | $0 |
| 5 hours / month | ~$1.50 | $15.00 | $0 |
| 20 hours / month | ~$6.00 | $15.00 | $0 |
| Accuracy | OpenAI gpt-4o-transcribe | Proprietary | Apple Speech |
| Works offline | No | Partial | Yes |
| AI enhancement | Yes (optional) | Yes | No |
| Multilingual | 10 languages | 100+ | Depends on Mac locale |
| Privacy | Key in Keychain, no servers | Cloud processing | On-device |
Prices reflect OpenAI's list rates as of April 2026. Actual costs depend on how much you dictate. You can set hard spending limits directly in your OpenAI dashboard.
Questions, answered
Do I need an OpenAI account?
Yes. ParlaParla uses OpenAI's API for transcription. You can create a free account at platform.openai.com — new accounts get a small free credit to test the app.
How much does it really cost?
OpenAI charges around $0.006 per minute of audio. That works out to roughly $0.30 per hour of dictation. Most users spend less than $2 per month.
Is my audio stored anywhere?
No. Audio is sent directly from your Mac to OpenAI's API and then deleted. OpenAI's own data policy applies — they don't train on API data by default.
Does it work without internet?
No. Transcription requires an internet connection to reach OpenAI.
Which Mac do I need?
macOS 15 Sequoia or later, Apple Silicon or Intel.
Can I use a different hotkey?
Yes. The default is Fn, but you can bind any shortcut you want in Settings.
What languages are supported?
Ten spoken languages for transcription: English, Danish, Norwegian, Swedish, German, French, Spanish, Italian, Dutch, and Portuguese. The app's interface is translated to eight of those.
Is ParlaParla open source?
Not currently, but we may open-source it in the future.