Skip to content
Lexena
v3.0 · public beta

Speak.
Lexena writes.

Hold your shortcut, speak out loud, the text appears exactly where your cursor is. Local or cloud — your call.

Local on GPU·Open source·Windows·MIT licensed
Target window
Recording…
 
Hold+F8
00:12
Live preview — dictation typing into a notes window
How it works

Three gestures. No friction.

No editor to open. No window to focus. Lexena lives in your tray and types into whatever you're already looking at.

  1. 01/03

    You press the key.

    Customizable global shortcut. Toggle, push-to-talk, or both. Works everywhere — Word, Slack, VS Code, browser, terminal.

  2. 02/03

    You speak.

    A floating mini-window shows the audio level and timer in real time. AI post-processing is opt-in and toggleable per dictation.

  3. 03/03

    The text appears.

    Auto-pasted into the active app instantly. A copy is kept in history with replayable audio — never without your permission.

For whom

Three usages, one app.

Write fast

Email, reports, drafts.

You speak faster than you type. Lexena reframes on the fly if you want — pro tone, email tone, bullet list, summary.

Made for writers, freelancers, consultants.

Tech & code

Pull requests, tickets, snippets.

Local Whisper on GPU so nothing leaves your machine. Configurable hotkeys, personal dictionary, snippet expansion for the acronyms no model recognizes.

Built for devs and power users.

Second brain

Connected notes.

Rich TipTap editor, @ backlinks, task lists, code blocks. Watertight Personal/Work profiles. Voice-captured thoughts, keyboard-organized.

For researchers, journalists, the curious.

What changes

Four things that make a difference.

Lexena isn't just a transcriber. Here's what makes daily use actually work.

Local · Whisper

Everything stays on your machine, if you want.

GPU-accelerated whisper.cpp (Vulkan) with CPU fallback. No connection required. No data leaves. No per-minute cost.
$0·0 ms network·0 leak
Coût mensuel0 €
Latence0,3 s
Données externes
Vulkan · GPU
CPU fallback
Modèle large multilingue
Post-process · AI

Reframed without becoming unreadable.

Ready-made modes: auto, email, formal, casual, bullet list, summary, grammar fix. Pick one per dictation. Original text always available.
Auto · Email · Formal·Casual · List · Summary·Raw kept
Post-process · Mode mail$0.0021 · 12s
Brut

« euh oui Marc je voulais te dire que le truc dont on a parlé hier c'est ok pour moi on peut y aller dis-moi quand »

Reformulé

Bonjour Marc, je donne mon accord sur le sujet évoqué hier. Indique-moi quand tu souhaites avancer.

AutoMailFormelCasualListeRésumé
Notes · TipTap

Your transcriptions become notes.

Full rich editor: headings, code blocks, tables, tasks, images. @ backlinks between notes. Full-text search. Synced across devices, free.
@ backlinks·Tasks·Free sync
Profiles · Personal & Work

A separate life per context.

Each profile has its own notes, history, hotkeys, dictionary. Instant switching. No leak from one context to another.
Watertight notes·Own hotkeys·Switch in 1 click
P
Perso
42 notes · 3 dictées · 18 min
dormant
P
Pro
184 notes · 12 dictées · 1h 04 min
actif
Ctrl+Shift+PBascule profil
Local or cloud

Both modes. No dogma.

We don't push you toward either. Pick per dictation, or let auto-mode decide based on context.

Compare in detail →
Local

Whisper on your GPU.

  • Fully offline — no data leaves the machine.
  • Vulkan (AMD/NVIDIA/Intel GPU) with CPU fallback.
  • Multilingual large model bundled.
  • ~0.3s latency on a recent GPU.
Free forever
Cloud · Pro

Lexena Cloud.

  • Lexena's cloud transcription, more accurate on tough audio.
  • Premium AI post-processing — extended modes.
  • Same data ethics: nothing stored server-side.
  • ~0.8s latency, hardware-independent.
TBD
Final pricing announced at v3.0 public launch
Get started

Ready to try?

Public beta available. Windows 10 or higher. Account creation in-app only — never on this site.

Lexena — Voice dictation for Windows