Hold. Speak.Polished.

Hold a shortcut, speak naturally, and clean text lands wherever your cursor sits.

uh, quick note tomorrow at three make it warmer actually scratch that client-ready new paragraph for the board deck maybe soften this Slack reply no, say Friday invoice follow-up add the context make it shorter legal tone Cursor note French greeting subject line clean it up um, I think I think dictionaire dictionary

Hi Arnaud, can we move tomorrow's call to 3 PM?

Voxa transforms

From ramble to ready.

Voxa turns messy spoken thoughts into clean, polished writing you can send with confidence.

uh quick note actually make that shorter for tomorrow's client call

Client call tomorrow

Just circling back to confirm our call tomorrow at 3:00 PM.

I'll have the latest updates and next steps ready to go.

Looking forward to it!

Polished

Email Before After

You say

hi just following up on the deck I sent yesterday um slide seven has a bit too much could we cut that section and move our call to four pm thanks

Voxa writes

Hi! Just following up on the deck I sent yesterday.

Slide 7 has a bit too much; could we cut that section?

Also, can we move our call to 4:00 PM?

Thanks!

Slack Before After

You say

heads up the deploy got pushed to tomorrow morning because of an edge case in payments uh nothing blocking I'll update in the thread

Voxa writes

Quick update: we're moving the deploy to tomorrow morning due to an edge case in payments.

Nothing blocking. I'll share updates in the thread.

Linear Before After

You say

add a task for the mobile onboarding flow uh we need analytics and an A B test let's target next Friday

Voxa writes

Title: Mobile onboarding flow

Description: Implement analytics events and A/B test for the headline.

Target: Next Friday

How it works

Three seconds. Three steps.

Hold a shortcut, speak naturally, and polished text lands wherever your cursor sits. The whole loop runs in under three seconds.

Ctrl + Shift + Space

01.

Hold the shortcut.

Default is Ctrl, Shift, and Space together. Change it in Settings if it collides with another app.

02.

Speak naturally.

Up to three minutes per take. Pause, change your mind, correct yourself. Say "new line" or "period" when you want them inserted exactly. The widget pulses with your voice while you talk and shimmers while it polishes.

Hi Arnaud, can we move tomorrow's call to three o'clock?

03.

Polished text lands.

Voxtral transcribes. Mistral cleans up filler, fixes punctuation, resolves self-corrections. The result pastes into whatever app the cursor was in.

See it in action

Dictate straight into Slack. Dictate straight into Email. Dictate straight into Notion. Dictate straight into Cursor. Dictate straight into ChatGPT. Dictate straight into Linear.

Voxa works where you write. Use the same shortcut to dictate in Slack, Email, Notion, Cursor, ChatGPT, Linear, and everywhere else.

See Voxa in action across your tools

Switch app demo

# design

A P M 5

A

Anna 10:42 AM

Hey team, can we shift the design sync to three?

P

Pim 10:43 AM

Three works for me. What about you?

M

Maya 10:44 AM

Yep, 3pm is good 👍

Y

You 10:45 AM

Three o'clock is good for me. I'll send a fresh invite now.

New

T

Tyler 10:46 AM

Let's also review the new onboarding flow.

Message #design Three o'clock is good for me. I'll send a fresh invite now.

Gmail

Search mail

? J

*

Sarah Thompson Project kickoff next steps

Hi team, thanks for a great kickoff today...

11:30 AM

*

David Lee Design feedback

Thanks for sharing the mockups...

10:15 AM

*

Marketing Team Q2 campaign update

Here's a quick update on our progress...

May 16

*

Jennifer Park Re: Client proposal

Looks good! A few small suggestions...

May 16

*

Product Updates What's new in May

This month we're excited to introduce...

May 15

Project update for Q2 - ↗ x

To Alex Morgan <alex.morgan@acme.com> Cc Bcc

Hi Alex,

Here's a quick update on our progress this quarter.

We've shipped the new onboarding flow, improved performance across key pages, and are seeing strong early feedback from customers.

Next up, we'll be focusing on scaling our integrations and refining the analytics dashboard.

Let me know if you'd like a deeper dive into any of these areas.

Thanks,
Jamie

Share [] ↻ ☆ ...

Weekly Team Sync

Date: May 12, 2024
Team: Product
Attendees: 5 people
+: Add a property

Agenda

Product updates
Roadmap review
Blockers

Notes

Capture the action items and send the summary to the product team after the sync.

task-manager

TS useTasks.ts x

src > hooks > TS useTasks.ts > useTasks

1234567891011121314151617181920212223242526

import { useState, useEffect } from 'react'
import { fetchTasks } from '@/lib/api'

export function useTasks() {
  const [tasks, setTasks] = useState<Task[]>([])
  const [loading, setLoading] = useState(true)
  const [error, setError] = useState<Error | null>(null)

  useEffect(() => {
    let cancelled = false
    const load = async () => {
      try {
        const data = await fetchTasks()
        if (!cancelled) setTasks(data)
      } catch (err) {
        if (!cancelled) setError(err as Error)
      } finally {
        if (!cancelled) setLoading(false)
      }
    }
    load()
    return () => { cancelled = true }
  }, [])

  return { tasks, loading, error }
}

CHAT COMPOSER + x

MAdd a new function to useTasks that creates a task.

Cursor

I'll add a createTask function that sends a POST request and updates the local state.

TS useTasks.ts +24 -0

export function useTasks() {
  // ...existing code...
  const createTask = async (title: string) => {
    const newTask = await fetch('/tasks', {
      method: 'POST',
      body: JSON.stringify({ title }),
    })
    setTasks(prev => [newTask, ...prev])
  }
}

ChatGPT 4o TS ...

Help me draft a launch plan for our new product.

Absolutely, happy to help. Here's a draft launch plan tailored to your new product.

Product Launch Plan

1. Goals

Generate awareness and excitement

Linear — Web

VOXA-148

In Progress Medium T Tim Cycle 12

See it in action: tabbed multi-app demo

Replace the single Slack mockup with a six-app tabbed demo. Each tab is a realistic UI for the app, with the Voxa capsule overlaying the dictation slot.

✓ Build SlackFrame
✓ Build GmailFrame
✓ Build NotionFrame
○ Wire auto-cycle JS

T

Timnow

Pushed the tabbed demo branch. Auto-cycle is 8 seconds per tab, manual clicks reset the timer. Ready for review.

Leave a comment... Pushed the tabbed demo branch. Auto-cycle is 8 seconds per tab, manual clicks reset the timer. Ready for review.

Your shortcut. Your flow. Works everywhere.

The wedge

Your voice never leaves Europe.

Most dictation tools route your audio through Silicon Valley. We do not. Voxa is engineered around three invariants that hold from the first ms of the first request.

Voxa API · Frankfurt Mistral models · Paris

01.

EU-only inference

Voxtral and Mistral models, both hosted on EU endpoints. Every request to transcribe and polish your dictation routes through Frankfurt.

02.

OpenAI Anthropic Google

No US AI subprocessors

No OpenAI. No Anthropic. No Google. The wedge is not marketing language, it is the architecture. Read the subprocessor list.

03.

Audio never persisted

Your voice streams straight into Voxtral and is discarded on response. No transcripts saved server-side, no final text retained.

Read the full privacy architecture

Built by

A Belgian developer who got tired of Silicon Valley listening in on his client conversations.

Tim Pieters · Brussels, 2026

I write code for a living and I dictate everything. Until last year I was happy with Wispr Flow. Then I read the architecture and realised my client conversations were round-tripping through Silicon Valley. For a Belgian developer working with European clients, that is not a thing you keep doing.

So I rebuilt it the way it should exist. EU-only inference, audio discarded on response, no transcripts retained server-side. Same hold-to-speak workflow, different geography under the hood.

Voxa is for the people who handle other people's data. Lawyers, consultants, devs, anyone who would rather not narrate their client calls into a US-hosted API.

Made in Brussels·Hosted in Frankfurt

Speaks your language

Six European languages. Auto-detected.

Voxtral identifies the language from the first words. Bias the detection by setting your UI locale in Settings if you switch a lot.

NL

Dutch

“Hallo”
FR

French

“Bonjour”
EN

English

“Hello”
DE

German

“Hallo”
IT

Italian

“Ciao”
ES

Spanish

“Hola”

Voxa

Two minutes.
Yours from Frankfurt.

Free for 2,000 words a week. Pro is €12 / month. No card to start, no signup form, no audio leaving Europe.

Download for Windows The privacy architecture

Hold. Speak.Polished.

From ramble to ready.

Three seconds. Three steps.

Hold the shortcut.

Speak naturally.

Polished text lands.

Dictate straight into Slack. Dictate straight into Email. Dictate straight into Notion. Dictate straight into Cursor. Dictate straight into ChatGPT. Dictate straight into Linear.

Weekly Team Sync

Agenda

Notes

Product Launch Plan

1. Goals

See it in action: tabbed multi-app demo

Your voice never leaves Europe.

EU-only inference

No US AI subprocessors

Audio never persisted

Six European languages. Auto-detected.

Two minutes.Yours from Frankfurt.

Two minutes.
Yours from Frankfurt.