A dictionary that builds itself
Quick-add any highlighted text with a hotkey, extract jargon from your documents with AI, or let it learn from your own corrections.
Meet Vox · your voice command deck
Talk. It types. It executes.
Hold a key, and just talk to me.
hold Alt+Space · talk anywhere
transcribingpasted · clipboard restored
Message #launch
Free tier included · No credit card · Other platforms
Three things no other dictation app has
Voice agents
Speak from any app → your n8n / Make / webhook fires → the reply lands in a popup. No competitor ships this.
True local mode
Transcription and AI cleanup run 100% on-device. Flip the switch and your voice physically cannot leave the machine.
Quick picker
Hold a key, flick your mouse, every voice tool in a radial under your cursor, without ever leaving the app.
since you landed, you could’ve spoken 0 words · typed: 0.
220 wpm
you speak vs ~45 typed
4–5×
faster than typing
100+
languages
~1s
local transcription on GPU tiers
100%
on-device option, by architecture
4
platforms · mobile in beta
[01] raw → polished
You ramble. It ships your tone.
Speak the way you actually think. Voxor strips the filler, fixes the grammar, and styles the result for wherever your cursor is, same words in, different register out.
raw transcript
uh okay so I was talking to the team about the launch and like we basically agreed that we need to push it to next tuesday because the api integration isn’t done and also marketing still needs the final copy you know and I think we should do a dry run friday
60 words · 6 fillers · one breath
styled output
30 words · casual · warm
[02] voice agents · nobody else has this
Don’t just dictate.
Delegate.
Every n8n workflow, Make scenario, or HTTP webhook becomes a voice command. Speak from any app, Voxor pipes the transcript to your endpoint and shows the reply in a popup you can keep chatting with.
- 1Hold your agent’s hotkey, every agent gets its own, like Right Ctrl
- 2Say what you need. Voxor can attach the text you highlighted and a screenshot of the active window
- 3The reply lands in a popup, copy it, or ask a follow-up without re-triggering
n8nMakeZapierany HTTP webhookBGOS assistants
Context goes straight to your endpoint, screenshots and highlighted text are never routed through Voxor’s backend for webhook agents.
hold hotkey to trigger
ACTIVE AGENTS
unlimited
PER-AGENT HOTKEYS
yes
FOLLOW-UP CHAT
built in
[03] the engine switch
Private by architecture,
not by promise.
Competitors sell retention policies. Voxor ships an actual on-device stack: GPU-accelerated speech recognition plus a local AI engine for cleanup, translation and instruction mode. Flip the switch, your audio simply never leaves.
engine readout
enginecloud streaming
audio leaves deviceyes, TLS to voxor.ai
works offlineno
final textlands as you release the key
Hard guarantee in code: Local Mode never silently falls back to cloud, a missing model is an error, not an upload.
cloud engine, when you want maximum accuracy
Streaming pipeline
audio streams while you speak, the finished, polished line lands the instant you release the key
100+ languages
plus live voice translation between them
Screen context (opt-in)
Voxor can see the active window so names and jargon come out right, off by default
Automatic fallback
if streaming hiccups, batch transcription takes over, you never lose a take
incognito mode5 min1 hour24 hoursuntil off· history off, screenshots off, logging off. With a live countdown.
[04] quick picker · ultra
Hold a key. Flick your mouse.
Every voice tool you have, in a radial under your cursor, it never steals focus from the app you’re in. Tap the same key instead and you get a searchable command palette.
hold Q
+ flick
Instruct
MODE · fixed slot
Speak a command instead of dictating, rewrite, reply, transform whatever you highlighted.
tap Q instead → command palette
trans
Translate, speak one language, paste another
Transcribe with Local engine
Ops Agent, “transfer the ticket…”
[05] instruction mode
Highlight. Speak. Done.
A separate hotkey for “do this” instead of “type this”. Voxor grabs whatever you’ve highlighted as context, runs your spoken instruction through an LLM, cloud or fully local, and pastes the result right where your cursor is.
Works in any appCloud or local LLMUses your highlight as context
Hi Dana,
Thanks for sending the revised scope over. I think we probably can’t really commit to the timeline you mentioned and honestly it might be better if we maybe push things back a bit. Happy to walk through the details on a call.
Best,
Sam
text highlighted · hold Ctrl+I and speak
[06] the rest of the deck
Power-user plumbing, everywhere.
Real screenshots, real features, this is the actual app, in both of its themes.
See exactly what the AI changed
History keeps a word-level diff between the raw transcript and the cleaned version, flip back to raw with one click, replay the audio anytime.
It types like you
Style Studio: six personas plus sliders for formality, warmth and density, with a live before/after preview. Set it once, every transcript matches.
Say “my address”, get the whole thing
Voice snippets expand trigger phrases into full saved text: addresses, sign-offs, boilerplate. Synced across your devices.
App-aware tone
Slack gets “hey”. Email gets “Dear”. Automatically, based on where your cursor is.
Voice translation
Think in Arabic, type in English. 100+ languages, translation runs offline in Local Mode too.
Deep context (opt-in)
Voxor can see your screen, only when you allow it, so it spells your client’s name right the first time.
Crash-proof recording
Audio is journaled to disk every second. Crash mid-thought? Recover and transcribe on relaunch.
Clipboard-safe paste
Types where your cursor is, then restores your clipboard, text and images. Win+V history stays clean.
Audio ducking
Spotify ducks itself when you start talking. Music and calls never bleed into your dictation.
[07] works everywhere you type
If there’s a cursor, Voxor types into it.
S Slack G Gmail N Notion V VS Code C Cursor C Chrome O Outlook W Word T Teams D Discord T Telegram W WhatsApp F Figma L Linear
O Obsidian C ChatGPT C Claude T Terminal E Excel S Safari J Jira D Docs A Arc i iMessage P Photoshop X Xcode R Reddit X X
[08] every device you own
Desktop today. Pocket next.
Windows and Mac are downloadable right now. The Android keyboard and iOS app are in beta, ask us for early access.
Windows
Windows 10/11 · x64 · NSIS installer
macOS
Apple Silicon · notarized DMG · Metal GPU
[09] the honest spec sheet
Same category. Different machine.
Everyone transcribes. Nobody else combines a true on-device stack with voice agents that act.
| capability | Voxor | Typical dictation apps |
|---|---|---|
| true local mode (win + mac) | Win + Mac | cloud-only |
| voice agents → webhooks | n8n / any HTTP | |
| hardware-matched local tiers | 4 tiers · auto-scan | |
| instruction mode | any app | ~ limited / add-on |
| screen context | opt-in | |
| win · mac · android · ios | mobile in beta | ~ 1-2 platforms |
| free tier | 2,000 wd/wk | ~2,000 wd/wk |
| pro price | $15 · $12 annual | $8-15/mo |
compared to typical cloud dictation apps · wd = words
[10] pricing
Start free. Scale when it sticks.
AnnualSAVE 18%+
Free
Try the whole loop, talk, polish, paste.
$0/mo
forever
- 2,000 words / week
- Basic AI editing
- Community support
MOST POPULAR
Pro
Unlimited dictation, tuned to your voice.
$12/mo
billed annually
- Unlimited words
- Advanced AI editing
- 100+ languages
- Custom dictionary
- Snippets
- Tone per app
- Priority support
Ultra
The full command deck.
$18/mo
billed annually
- Everything in Pro
- Voice agents
- Quick picker
- Shared dictionaries
- Dedicated support
Prices in USD. Free tier needs no credit card. Plans manage at portal.voxor.ai.
[11] straight answers
Questions, answered honestly.
Does Voxor work offline?
Yes, turn on Local Mode and transcription runs fully on-device, with GPU acceleration (CUDA on Windows NVIDIA, Metal on Apple Silicon). The optional local AI engine also runs cleanup, translation and instruction mode offline. Word quotas check in with a 24-hour offline grace window, so a flight never bricks you.
Is my voice actually private?
In Local Mode, audio never leaves your machine, the code refuses to fall back to cloud rather than upload. In cloud mode, audio streams over TLS to voxor.ai for transcription. Incognito Mode adds timed toggles (5 min / 1 h / 24 h) that stop history, screenshots and logging. Screen context is off by default.
What hardware do I need for Local Mode?
The Basic tier (a 75 MB model) runs on almost any laptop with 4 GB of RAM. Tiers scale up to Ultra for machines with 24 GB+ VRAM or higher-end Apple Silicon. Voxor scans your CPU, RAM and true GPU VRAM and recommends the right tier, you can override it anytime.
What do I get on the free tier?
2,000 words per week, basic AI editing, and community support. No credit card needed. Upgrade to Pro for unlimited words, or Ultra for voice agents and the quick picker.
Which platforms are supported?
Windows 10/11 (x64) and macOS (Apple Silicon) are downloadable today. The Android voice keyboard and the iOS app are in beta, email support@voxor.ai for early access.
How do voice agents work?
Give any agent its own hotkey and point it at an HTTP webhook (n8n, Make, Zapier, your own server) or a BGOS assistant. Hold the hotkey, speak, and Voxor sends the transcript, optionally with your highlighted text and a screenshot of the active window, to your endpoint, then shows the reply in a popup you can keep chatting with. For webhook agents, that context goes only to your endpoint, never through Voxor’s backend.
Will it paste into any app?
Yes, Voxor types where your cursor is in any app, then restores whatever was on your clipboard (text or images). There’s even a mode to keep transcripts out of your clipboard history.
Your keyboard had
a good run.
Two minutes to install. One hotkey to learn. Every app you already use.
Free tier included · No credit card · Other platforms