Aqua Voice - AI Dictation for Mac, Windows, and iPhone

TL;DR

TL;DR: Aqua Voice is a YC W24-backed desktop and mobile dictation app that uses a proprietary Avalon transcription model to deliver sub-50ms startup, 97.3% accuracy on the AISpeak benchmark, natural-language editing commands, and 49-language dictation across Cursor, Slack, Gmail, ChatGPT, and any text field on your system.

Source and Accuracy Notes

• Project page: withaqua.com
• Official pricing: aquavoice.com/pricing
• App Store: apps.apple.com/us/app/aqua-voice/id6759074969
• HN Launch thread (716 pts): news.ycombinator.com/item?id=39828686
• HN Show HN thread (140 pts): news.ycombinator.com/item?id=43634005
• Pricing verified from aquavoice.com/pricing: Pro $8/mo billed annually ($96/year) or $10/mo month-to-month; Max $24/mo annually ($288/year) or $30/mo monthly; Team $12/user/mo billed annually; Business custom; .edu address gets 70% off Pro and Max.
• Language count (49) and accuracy (97.3% AISpeak, sub-50ms / ~450ms latency, SOC 2 Type II) verified from the App Store listing description.

Affiliate disclosure

runany.dev is enrolled in Aqua Voice’s creator program via Dub Partners. The runany.dev tracking link is aquavoice.dub.link/runany. If you sign up through that link, runany.dev earns:

• 25% commission on every paid plan sale for 1 year from the signup date.
• 25% commission on every referred partner’s commission for 1 year.
• $10 off your first month on Pro or Max (the new-user discount above).

The technical content of this post is independent of the affiliate relationship — pricing, language count, and accuracy numbers above were verified against aquavoice.com/pricing and the App Store listing, not against Aqua Voice’s marketing material.

What Is Aqua Voice?

Aqua Voice is a voice-driven text input tool built by YC W24 founders Finn and Jack. Finn has been using dictation software since sixth grade to manage dyslexia, and the pair built Aqua after years of frustration with existing tools that focus on raw transcription accuracy but ignore the real problem: turning messy speech into clean, correctly formatted text.

The product ships as a desktop app for Mac and Windows, plus an iPhone app (iOS 17.0+, 42.7 MB). It sits as a system-wide input method, so you can dictate into any text field — code editors, email clients, chat apps, terminal prompts — without switching context.

How Aqua Voice Works

Aqua runs six models in parallel to handle the full pipeline from audio to formatted text:

1. MoE transcription — a mixture-of-experts speech model handles real-time audio-to-text
2. Language model sync — determines what should appear on screen as quickly as possible
3. Intent classification — decides whether your speech is content (transcribe verbatim), a command (execute it), or a rough draft (clean it up)
4. Formatting pass — applies context-aware formatting based on the target app

The result, per the founders’ HN post:

• Startup in under 50ms
• Text insertion in about 1 second (sometimes as fast as 450ms)
• 97.3% accuracy on the AISpeak benchmark (technical vocabulary — API endpoints, model names, programming terms)

Setup and Installation

Mac / Windows Desktop

Download from withaqua.com/download or aquavoice.com/download. The app installs as a system-wide input method.

iPhone

Install from the App Store (requires iOS 17.0+).

Using Aqua Voice

Once installed, activate dictation with a shortcut (configurable). Speak naturally — Aqua handles punctuation, filler-word removal, and formatting. You can also issue editing commands:

"Remove filler words, this is a slack message"
"Translate this to Japanese"
"Make this title case"
"Spell Toni T-O-N-I"
"Format the numbers like $10m"

Aqua works inside any app that accepts text input: Cursor, VS Code, Slack, Gmail, ChatGPT, Claude, terminal, and more.

Pricing

Aqua Voice’s free tier starts you off with 1,000 words on the Avalon transcription model. Paid plans are billed through aquavoice.com/pricing and the in-app checkout, with annual billing giving the better per-month rate:

Plan	Annual (per month)	Annual total	Month-to-month
Free	$0	$0	$0 (1,000 words on Avalon)
Pro	$8/mo	$96/year	$10/mo
Max	$24/mo	$288/year	$30/mo
Team	$12 per user/mo	—	—
Business	Custom	—	—

What’s in each tier:

• Pro — Unlimited words, Custom Instructions, expanded Custom Dictionary.
• Max — Everything in Pro plus Realtime Mode, the “Send it” voice command, and early access to new features.
• Team — $12 per user/month billed annually, for shared workspaces.
• Business — Custom-priced for larger deployments.

Students and educators with a .edu email address get 70% off Pro and Max — Pro lands at $2.40/mo billed annually ($3/mo month-to-month), Max at $7.20/mo annually ($9/mo month-to-month).

New users signing up through the runany.dev referral link also get $10 off their first month on Pro or Max — enough to cover most of a monthly Pro subscription as a free trial: aquavoice.dub.link/runany.

Supported Languages

Aqua Voice dictation covers 49 languages, including English, Spanish, French, German, Italian, Portuguese, Dutch, Polish, Russian, Ukrainian, Greek, Turkish, Arabic, Hebrew, Hindi, Bengali, Tamil, Thai, Vietnamese, Indonesian, Malay, Filipino, Japanese, Korean, Mandarin Chinese (Simplified and Traditional), Cantonese, and many more. The full list lives at aquavoice.com/pricing under “Languages”.

Privacy and Data Handling

Aqua Voice is not a fully local/offline tool. The product uses cloud processing for transcription, and the privacy policy references sharing data with service providers including OpenAI. The founders have been transparent about this in HN comments.

If you need fully local, offline dictation, consider open-source alternatives like VoiceInk or whisper-dictation.

Aqua maintains a Vanta trust page for security compliance details, and the iOS app advertises SOC 2 Type II certification on its App Store listing.

FAQ

Q: Is Aqua Voice open source? A: No. Aqua Voice is a proprietary product. The founders have not announced plans to open-source any part of the platform.

Q: Does Aqua Voice work offline? A: No. Aqua Voice requires an internet connection for cloud-based transcription. The product uses OpenAI as a service provider for some processing.

Q: How does Aqua Voice compare to Apple Dictation or Whisper? A: Aqua Voice focuses on turning messy speech into clean, formatted text with natural-language editing commands. Apple Dictation and raw Whisper models handle transcription but don’t distinguish between content and commands, and don’t auto-format output.

Q: Can I use Aqua Voice with ChatGPT or Claude? A: Yes. Aqua Voice works as a system-wide input method, so you can dictate directly into ChatGPT, Claude, or any other text field on your system.

Q: What platforms are supported? A: Mac, Windows, and iPhone (iOS 17.0+). The founders have mentioned plans to support Linux.

Conclusion

Aqua Voice fills a real gap in the dictation market by treating voice as a first-class input method rather than just a transcription engine. The combination of sub-50ms startup, 97.3% accuracy on the AISpeak benchmark, 49-language coverage, and natural-language editing commands makes it viable for developers, writers, and anyone who thinks faster than they type. The cloud dependency is a trade-off — you get speed and accuracy but lose offline capability and full data privacy. Pro starts at $8/mo billed annually ($10/mo month-to-month), and the free 1,000-word tier is enough to evaluate it before committing.