Context-Aware Voice Typing for Mac

Speak. It already knows

Typing makes you compress every idea to fit your finger speed. Loqua gives your voice the full bandwidth — context-aware, scene-intelligent, already formatted when it arrives

Up to 4x faster than typing. With context that makes every word count

  • Works in Any App
  • Context-Aware AI
  • Zero Copy-Pasting

The gap between speaking and sending, closed

Most tools transcribe. Loqua understands — the app you're in, the context you're working in, and what the output needs to look like when it gets there

Input

One shortcut. Every app. Every language

Press, talk, keep going. Loqua activates system-wide — no per-app setup, no mode switching. Dictate in VS Code, Slack, Mail, Notes, or your novel draft. Switch between English and Chinese mid-sentence. Your cursor is the only context you need to set

VS Code Slack Mail Notes EN / 中
Clean

Say it messy. It comes out right

Real speech has false starts, filler words, and mid-sentence second thoughts. Loqua strips what you didn't mean and keeps what you did — including the self-correction you made three words in. What arrives is clean enough to send without touching it

You said

um so basically the thing is we need to uh... actually wait let me start over — we should cache this for like 15 minutes

What arrived

We should cache this for 15 minutes

Understand

It reads the room

The same idea needs to land differently in VS Code than in iMessage. Loqua detects the app, the register, and the recipient — and adjusts the output without you asking. The more you use it, the more invisible it becomes

You say
"let them know Thursday works, I can do morning or afternoon, whatever's better"
Gmail — New Message
Output

Right where your cursor is

No copy-paste. No app switching. Text appears at the cursor, replaces selected content, or queues to clipboard — without breaking what you were doing. Nothing to copy. Nothing to paste. Just what you said, already there

Gmail — Reply

Hi Sarah,

Thanks for sending the proposal over

Try it now

Toggle Ctrl to start speaking, then toggle Ctrl again to stop and finish the transcription.
Your words appear here, formatted and ready.

Loqua — Demo
Press Ctrl to start speaking

Three models. One shortcut

Speech recognition Language intelligence Multimodal context

A purpose-built stack. Not an API wrapper. Not a thin client

Named Entity Recognition

Your technical vocabulary, product names, teammates, APIs, and recurring jargon — detected and handled automatically. The common cases don't need a dictionary entry. The uncommon ones learn fast

Structured Output

What arrives isn't a transcript you fix. It's formatted text that matches the expected shape of the surface — bullet lists in Notes, code comments in VS Code, professional paragraphs in email

Screen Context

Loqua reads the active app, adjacent text, and selected content to understand where your words are going and what they should look like when they get there. Same input, different result

For every register of your day

Loqua was built by a team that writes code, commit messages, Slack threads, client emails, product briefs, and the occasional chapter. It shows

Coding

The insight is in your head. Don't stop to type it

You say

"add a comment saying we cache this for 15 minutes and if the refresh token's expired just redirect to login don't retry"

VS Code
// Cache duration: 15 min // On token expiry (401): redirect to /login — do not retry
You say

"this fixes the race condition in session cleanup when two logout requests hit at the same time, added a mutex on the session store, also added a regression test"

GitHub PR
fix(auth): handle concurrent session cleanup race condition Two simultaneous /logout requests could both attempt to invalidate the same token. Added mutex lock on session store write path. Tests: auth_test.go — TestConcurrentLogout
You say

"okay train a resnet50 on the augmented dataset use cosine annealing with warmup and log to wandb, also early stop if val loss doesn't improve for 5 epochs"

Cursor
# Train ResNet-50 on augmented dataset # LR schedule: cosine annealing w/ warmup # Logging: W&B # Early stopping: patience=5 on val_loss
Life

Walking out the door. Three things to send. Zero time to type

You say

"running 15 late start without me I'll grab coffee"

iMessage
Running about 15 minutes late — go ahead and start. I'll grab coffee on the way
You say

"okay what if Elena already knew about the letter this whole time and has been waiting for someone to bring it up"

Notes

Chapter idea — Elena already knows about the letter. Has been waiting for someone to mention it. Reframes her silence through Act 2: not passive, deliberate

You say

"yeah dinner sounds good how about that ramen place near the station like 7ish"

WhatsApp
Dinner sounds great! How about that ramen place near the station — around 7?
Work

Four minutes between meetings. Six messages to send

You say

"hey team staging is back up QA can restart the checkout flow testing we're good to go"

Slack

Staging is back up — QA team, go ahead and restart checkout flow testing. We're good. ✅

You say

"tell them I loved the conversation and I want to move forward with the pilot, can we find a time next week"

Gmail

Loved our conversation — I'd like to move forward with the pilot. Would you be available for a call next week? Tuesday or Wednesday would work best on my end

You say

"okay so I need to compare our three enterprise tiers, look at last quarter's conversion data, figure out which one we should kill, and give me a migration plan for existing customers on that tier"

ChatGPT

Compare our three enterprise pricing tiers using Q4 conversion data. Specifically:

  • Revenue per tier vs. acquisition cost
  • Trial-to-paid conversion rate by tier
  • Churn rate and expansion revenue
  • Recommendation: which tier to deprecate and why
  • Migration plan for existing customers on the deprecated tier

Used by people who can't afford for it to be wrong

"I narrate my PR descriptions while I'm still looking at the diff. Loqua handles the format. I ship faster and the descriptions are actually better than what I used to write"

James Chen
Staff Engineer — Product Hunt

"I plot out loud. Loqua catches the good parts and discards the noise. It's changed how I draft — I get more on the page and edit less afterward"

Sarah Liu
Fiction Writer — Mac App Store

"Back-to-back meetings all day. Four minutes between each one. Loqua is how I actually get follow-ups out. Nothing gets dropped anymore"

Marcus Park
Startup Founder — X

"It handles technical vocabulary — model names, dataset formats, citation styles — better than anything I've tried. I stopped correcting output weeks ago"

Yuki Tanaka
ML Researcher — Product Hunt

"My thinking is faster than my typing. Always has been. Loqua is the first tool that doesn't penalize me for that"

Priya Mehta
PM — Mac App Store

"I dictate documentation, READMEs, and talk scripts. It sounds like I wrote them carefully. I didn't"

David Torres
Developer Advocate — X

Start free. Upgrade when you can't stop

No credit card upfront. Use the core product free — upgrade to Pro when Loqua is already part of your day

Basic
Free
No credit card required

Try the engine. Core dictation, smart cleanup, basic context detection.

  • Core dictation
  • Smart cleanup
  • Basic context detection
  • 5 dictionary entries
Get started
Pro Monthly
$12/mo
Billed monthly

The full thing: unlimited dictation, context intelligence, and more.

  • Unlimited dictation
  • Full context intelligence
  • Multimodal screen reading
  • Full personal dictionary
  • History & priority support
Get Pro Monthly

Building the future? It’s on us.

We support the open-source community. Maintainers of public OSS projects can apply for a complimentary lifetime Pro license.

Apply for Developer Grant

Your thinking, uncompressed

Download Loqua and start speaking. It's free

Download for Mac