NEWIntroducing Avalon
Read the announcement

Speak naturally, ship faster

Pro voice input that saves you 30 minutes of typing every day

iMessage interface
Cursor interface
Bolt interface
Technical accuracy
Streaming
App Agnostic
Powered by Avalon
Y CombinatorBacked by Y Combinator

Trusted by developers at Google, OpenAI, and Anthropic

Ship faster with voice that actually works

Aqua Voice eliminates the friction between your thoughts and your code. Write complex prompts, document functions, and message teammates—all at the speed of speech. No context switching, no copy-pasting, no fixing autocorrect disasters.Thanks to Avalon—our model optimized for Human-AI interaction—Aqua Voice delivers 97% accuracy on technical terms. This means your AI prompts work the first time, your code comments are always correct, and you never waste time fixing "use effect" back to "useEffect". It's the difference between typing for 5 hours and shipping in 3.

Who uses Aqua?

From accessibility to creative workflows, professionals trust Aqua to save time every day
Accessibility
Consultants
Creators
Customer Support
Designers
Educators
Engineers
Executives
Government
Healthcare
Lawyers
Journalists
Product Managers
Sales Teams
Students
Writers

Instant Mode: Super fast. Minimal.

Instant
~450ms

Starts up in under 200ms. Pastes complete text in as little as 450ms.

Streaming Mode: Deep context. Powerful.

Streaming
~850ms

Responds in ~850ms with continuous output as you speak. Maintains deep contextual understanding throughout your conversation for complex tasks.

Works Everywhere You Do

Aqua Voice runs locally on your Mac or PC, bringing Avalon's accuracy to every application—from Cursor to Messages to Slack.

Used in 50,000+ coding sessions this week

iMessage interface
Powered by Avalon
Cursor interface
Bolt interface
Technical accuracy
Streaming
App Agnostic

Voice Input That Actually Understands Your Workflow

Aqua Voice saves you 30+ minutes of typing every day by getting technical terms right the first time. No more breaking your flow to fix "cube cuttle" → kubectl or "high torch" → PyTorch.

Aqua Voice
Grok 4 beats GPT-5 on Arc AGI.
Others
Brock 4 beats GPT-5 on ARK AGI.

Because Aqua Voice is powered by Avalon—a model trained on real developer conversations—your prompts are more accurate, your code comments are cleaner, and you spend zero time fixing transcription errors. This means better AI responses, fewer bugs from typos, and more time actually building.

While other voice tools struggle with technical vocabulary, Aqua Voice understands the difference between "npm install" and "NPM in stall," between "useState" and "use state." It's the difference between a tool that fights you and one that flows with you.

Stop wasting time on corrections. Download Aqua Voice and start shipping faster with voice input that actually works for developers.

Ready to stop typing and start talking?

Free to try · Works with all your apps

Use Aqua with your favorite apps

cursor
bolt-new-bg
bear
iawriter
claude
windsurf
vscode
No need for app specific plugins. Aqua works with existing workflows on apps, websites, and documents. On Windows and Mac.

Enhanced With Deep Context

Screen
Context

Enhanced
Understanding

Better
Responses

Improves accuracy for coding, messaging, and document editing by understanding what's on your screen. Your data is processed securely and never stored.

Context Awareness
Secure Processing

Global Language Support

Aqua supports 49 languages, providing accurate voice-to-text for users worldwide.

Note: Avalon is currently available for English only. Support for all languages coming in the next few weeks.

49 Languages
Industry-leading accuracy
ArabicBelarusianBengaliBulgarianCantoneseCatalanCroatianCzechDanishDutchEnglishEstonianFinnishFrenchGalicianGermanGreekHebrewHindiHungarianIndonesianIrishItalianJapaneseKoreanLatvianLithuanianMalayMalteseMandarinMarathiMongolianNorwegianPersianPolishPortugueseRomanianRussianSlovenianSpanishSwahiliSwedishTamilThaiTurkishUkrainianUrduVietnameseWelsh

Details without the Effort

siri logo

Siri

Email
hey me call can you please review the marketing proposal i sent over yesterday i think we need to focus on three main areas first the social media strategy needs more specific targeting second we should reconsider the budget allocation for q3 and third the timeline for the product launch seems too aggressive given our current resources let me know what you think thanks tom
aqua logo

Aqua

Email
Hi Mikal, Could you please review the marketing proposal I sent yesterday? I think we need to focus on three key areas: 1. Social media strategy needs more specific targeting 2. Budget allocation for Q3 should be reconsidered 3. Product launch timeline may be too aggressive for our resources Let me know your thoughts. Thanks, Tom

Syntax highlighting in your transcript (nice)

google logo

Google

Talking about Code
create a function called get user data that takes an ID parameter and returns a promise that resolves with the user object first we need to fetch from the API endpoint slash users slash ID then we convert the response to Jason and return it if there's an error we should log it to the console and throw a new error saying failed to fetch user data
aqua logo

Aqua

Talking about Code
Create a function called `getUserData` that takes an `ID` parameter and returns a promise that resolves with the `user` object. First we need to fetch from the API endpoint `/users/${id}` then we convert the response to JSON and return it. If there's an error we should log it to the console and throw a new error saying failed to fetch user data

doesn't text like a grandma

dragon logo

Dragon Dictation

iMessage (gen z style)
Hey, I'm going to be a little late to the party. I'll probably get there around 9:30 PM. Do you want me to bring anything? I could stop by the store on my way there. Let me know. Thanks!
aqua logo

Aqua

iMessage (gen z style)
hey gonna be a lil late to the party... prob get there around 9:30 lol want me to bring anything? can stop by the store otw lmk thx

97% Accurate on Tech Talk

We tested how well different models transcribe the way developers actually talk. Avalon gets terms like "GPT-4o", "kubectl", and "PyTorch" right.

Technical Term Recognition Accuracy

Higher is better · Based on 500+ developer recordings

Avalon
97.3%
AssemblyAI Best
71.2%
Whisper Large v3
65.1%
ElevenLabs Scribe
62.4%
NVIDIA Canary 1B
51.5%

Real examples from our testing:

Avalon
Let me make something plain. There's only one instance where a sushi belt should be used in Factorio, and that is when you're building a hauler spaceship.
Whisper
Let me make something plain. There's only one instance where a Sushi Belt should be used in Factorio: Gama, and that is when you're building a hull or spaceship.
Avalon
Deploy with kubectl
Whisper
Deploy with cube cuttle
Avalon
Build with PyTorch
Whisper
Build with high torch
Avalon
Update my zshrc
Whisper
Update my Z short C

More Accurate and Up to 31% Faster

Avalon outperforms other models on standard speech benchmarks while processing up to 31% faster.

Word Error Rate across benchmarks

Lower scores indicate better performance

Word Error Rate across benchmarks

Avalon wins on 7 of 8 standard benchmarks · Latency up to 31% faster on typical requests

Simple, transparent pricing

Avalon included in all plans • Launch week: Pro users get priority access to new features

Starter

Free
Perfect for getting started
  • 1,000 free words
  • Aqua Engine
  • 5 Custom Dictionary values
Popular

Pro

$10/mo
billed monthly
$8/mo
$96 billed annually
  • Unlimited words
  • Aqua Engine
  • Avalon model
  • 800 Custom Dictionary values
  • Tune with Custom Instructions

Team

$15/mo
billed monthly
$12/mo
$144 billed annually
  • Everything in Pro
  • Avalon model
  • Centralized team billing
  • Team wide settings
  • Enforce privacy mode org-wide

Start saving 30 minutes every day

Join thousands of developers who ship faster with Aqua Voice