Heario v0.1.1 just launched — Download free for Windows →
Heario v0.1.1 — Free for Windows
Heario mic icon

AI Copilot for
Interviews & Meetings

Heario takes perfect meeting notes and gives real-time answers, all while completely undetectable.

300msResponse time
95%Transcription accuracy
100%Undetectable
1+ Downloads — Live Tracker · Watch it go up as you become a Heario

See it in action

Four Ways We Make Your Meetings Better

One AI that listens, thinks, and helps — invisibly.

Core

Real-time Question Answering

The right answer, right when you need it — Heario listens, understands, and streams exactly what to say in under 300ms.

listening · S0
Q: How would you scale this to a million users?
Any model

Universal LLM Interface

Bring your own keys or use our managed plan.

AI
GPT Claude Gemini
Grounded

Profile Intelligence

Never generic — always you.

Résumé Job Description Custom Context Past Answers
Auto-captured

Structured Meeting Notes

Searchable, timestamped notes — action items, decisions and key moments, exported automatically at session end.

  • 00:14 Decision: adopt token-bucket rate limiting
  • 02:47 Action: benchmark Redis vs. in-memory
  • 05:03 Key moment: candidate proposed sharding
  • 08:21 Follow-up: send system-design diagram

✦ v0.1.1 · Just Shipped

Fourteen reasons to update

A whole new layer of control — type, switch, pin, hide, and go fully offline. Watch each one work, live.

listening Type to Ask Heario
How does a bloom filter work?
C5

A bloom filter is a space-efficient probabilistic set. It answers "definitely not present" or "possibly present" using a bit array and k hash functions — no false negatives, tunable false positives.

Typed by you · answered in 280 ms
Claude GPT-4o
claude-opus-4.8 gpt-4o

One interface, your choice of brain. Drop in either key — switch mid-session, no restart.

Provider · live swap
BriefNormalDetailed

Use a queue to decouple writes.

Use a queue to decouple writes from reads, so spikes drain smoothly instead of overwhelming the DB.

Use a queue to decouple writes from reads, so traffic spikes drain smoothly. Add idempotency keys, a dead-letter queue for poison messages, and back-pressure so producers slow when consumers lag.

One key cycles the depth
Interviewer speaking — answering You're speaking — holding the answer
Diarization-aware · never interrupts
00:14 Decision: token-bucket rate limiting 02:47 Action: benchmark Redis vs in-memory 05:03 Key moment: candidate proposed sharding
session.pdf session.txt
Exported automatically at session end
Stripe
  • Industry — payments infrastructure for the internet
  • Stage — late-stage private, ~$95B valuation
  • Recent — expanded into billing, tax & issuing
  • Values — rigor, low ego, "developers first"
Injected into your AI context
Cloud STT + LLM Local Whisper + Ollama
and that's how I'd scale it.
Optional fully-local mode (via Ollama) · zero data breaches · no internet required
Deepgramstreaming · diarized · $200 free creditselected
OpenAI Whispercloud · your keyselected
Local Whisperoffline · freeselected
Auto picks the best key you have — or override it
C5

Approach: queue the writes so spikes don't hit the DB directly.

Use a queue to decouple writes from reads, so traffic spikes drain smoothly instead of overwhelming the database.

Toggle on/off · says why before what
Stripe Anthropic Datadog
Job description Company brief Default mode Mode prompts
Save a whole setup per company — switch in one click
C5

Shard by tenant, cache hot reads in Redis, and put writes behind a queue.

Pin
Pinned 1 Shard by tenant, cache hot reads in Redis…
Keep your best answers one tap away — copy or revisit any time
Heario
Heario
Collapse the overlay to a tiny status pill — then expand right back
Ctrl + Shift + H
C5

Use a token bucket in Redis — INCR with a TTL per client key…

✓ Gone — instantly
A global hotkey that works even when Heario isn't focused
Questions Asked

How would you scale this to 10M requests/day?

How You Answered

Shard by tenant, cache hot reads, queue the writes.

Follow-Ups to Prepare

Rehearse cache-invalidation tradeoffs.

One click at session end — questions, answers & what to prep next

◇ Adaptive Intelligence

Seven Modes. One AI That Listens.

Switch live mid-session. No restart — the core just retunes.

listening 💻 Technical Interview Crisp answers to coding & system questions.

Hover a mode to retune the core · or press M in-app to cycle live

Completely Undetectable

No meeting bots. No screen recording. Nothing visible to the other side.

1

Loopback capture

Listens to what your speakers play — the interviewer's voice — via OS-level audio. No mic required.

2

Instant transcription

Streaming speech-to-text with diarization. Knows who's talking. Ignores crosstalk.

3

AI answers

Claude or GPT generates a tailored answer using your background. Streams in under a second.

4

Hidden overlay

An always-on-top window invisible to screen capture. Only you see it. Drag anywhere.

Invisible by design

They Can't See It.
Only You Can.

Heario uses a Windows API call — SetWindowDisplayAffinity — that tells the OS to exclude the overlay from all screen capture pipelines. Zoom, Teams, Google Meet, and OBS all see a blank space where Heario sits. This isn't a trick or a workaround; it's the same mechanism used by banks and DRM software to protect sensitive content.

  • No browser extensions to detect
  • No meeting bots joining your call
  • No screen recording of your answers
  • No network traffic visible to the interviewer
CAPTURE FEED · interviewer REC
SC
Sarah Chen
You
You
EXCLUDED FROM CAPTURE
YOUR DISPLAY · live VISIBLE
SC
Sarah Chen
You
You
Heario C5
Use a token-bucket in Redis — each client key gets INCR + a TTL in one Lua script, so the limit check stays atomic…

Common Questions

Everything you need to know before your next interview.

Will Zoom / Teams / Google Meet detect it?

No. Heario's overlay uses SetWindowDisplayAffinity(WDA_EXCLUDEFROMCAPTURE) — a Windows OS-level flag that excludes the window from all screen-capture pipelines before the frame ever reaches Zoom's encoder. The interviewer's screen share sees nothing. This is the same API used by banking apps to protect sensitive data on screen.

Does it join the call as a bot?

Never. Heario runs entirely on your local machine. There is no bot, no browser extension injected into the call, and no third-party service that touches your meeting. It listens to your speaker output using a standard Windows audio loopback — the same way recording software works — and the interviewer sees nothing unusual.

Do I need to sign up for anything to use Heario?

No. It works immediately with no account and no API key, using your computer's own Whisper speech-to-text.

Does it cost anything by default?

No. The built-in fallback (Local Whisper) is free and runs on your machine.

Is my audio sent to the cloud?

Not by default. With no keys configured, transcription happens locally. Answers (the AI's replies) still go to Claude or GPT's cloud by default — only the transcription is local out of the box.

The first time I open it, it says "downloading 150MB" — is that normal?

Yes. The first time Local Whisper runs, it downloads its speech model once. After that, it's cached and starts instantly every time.

Can I make the AI's answers fully local too, not just transcription?

Yes, but that part isn't automatic. You need to separately install a free program called Ollama on your computer, then point Heario at it. It's an optional, one-time setup — not something that happens by default.

If I add a Deepgram or OpenAI key, what changes?

Transcription switches from local Whisper to that cloud service instead — faster and more accurate, but no longer offline, and (for Deepgram) usage-based.

So what's actually "zero data breach risk"?

Only the fully-local combo (Local Whisper + Ollama) is 100% offline end-to-end. Local Whisper alone is offline for transcription only; if you're using Claude/GPT for answers, that part still leaves your machine.

How fast are the answers?

Deepgram's streaming STT delivers a transcript in under 300ms from when the interviewer finishes speaking. Claude then streams the answer token-by-token — you start reading within a second. Total time from question to first word on screen is typically 1–2 seconds in normal network conditions.

What if I'm asked a follow-up question?

Heario keeps a rolling conversation memory of the last 4 turns. So if the interviewer says "and how would you scale that?", the AI already knows what "that" refers to and gives a contextually grounded follow-up, not a generic answer.

Can I use my own OpenAI or Claude key?

Yes. Bring your own API keys for OpenAI (GPT-4o) or Anthropic (Claude) in the .env file and you pay only the model's API rate — no markup. Or use our managed plan and we handle keys, rate limits, and billing.

Does it work on Mac?

Currently Windows-only. The undetectable overlay relies on SetWindowDisplayAffinity, a Windows-exclusive API. Mac support is on the roadmap — join the Mac waitlist to get notified.

What does the Web Search button do?

When enabled, Heario searches the web before generating each answer, so Claude has access to current information — recent framework docs, company news, live pricing, anything that might be out of date in a model's training data. It works out of the box with no setup using DuckDuckGo. If you add a free Tavily API key in Settings, it automatically upgrades to higher-quality AI-optimised results. Toggle it on or off mid-session with the button or by pressing W.

What are the 7 modes and when should I use them?

Each mode tells Heario's AI exactly how to respond for a different situation. Technical Interview gives crisp answers to coding and system questions. Behavioral structures answers using the STAR method from your background. Sales handles objections and advances the deal in real time. Lecture captures key points as concise notes. Recruiting coaches the interviewer side — flagging weak candidate answers and suggesting follow-up questions. System Design provides architecture hints and scalability prompts. Mock Interview acts as a tough coach, pointing out weak answers and suggesting stronger phrasing. Switch between them instantly with the mode button or by pressing M.

Microphone or System Audio — which should I use?

Use Microphone for live video calls on Zoom, Teams, or Google Meet — it captures what's being said in the room or through your headset. Use System Audio to capture anything playing through your speakers, like a YouTube mock interview, a podcast, or a recorded call. You can switch between them in Settings → Audio Source and apply the change without restarting the app.

What Users Are Saying

From engineers who used it in real interviews.

★★★★★

"I used Heario for a system design round at a FAANG company. The answers were sharp, grounded, and streamed fast enough that I could read ahead while the interviewer was still talking. Got the offer."

MK
Marcus K.Senior SWE, hired at Meta
★★★★★

"Genuinely invisible. I tested it by screen-sharing to a colleague — they saw nothing. The overlay just disappears. The behavioral mode is perfect for those 'tell me about a time when...' curveballs."

SR
Sophie R.Product Manager, London
★★★★★

"The latency is what sold me. Other tools I tried had a 3–4 second lag which made them unusable in a real conversation. Heario is there before I've even finished processing the question myself."

JL
James L.Backend Engineer, Berlin

Simple Pricing

Start free. No card required.

Managed Plans · Lifetime / Yearly
Free Trial
$0/mo
No card required
  • ✓ 10 AI requests
  • ✓ 10 min speech-to-text
  • ✓ 2 web searches
  • ✓ All 7 modes
  • ✓ 1 Quick Debrief
Download Free
Standard
$8/mo
For regular users
  • ✓ 500 AI requests
  • ✓ 200 min speech-to-text
  • ✓ 20 web searches
  • ✓ All 7 modes
  • ✓ Unlimited meeting debriefs
Get Standard
Max
$24/mo
For heavy users
  • ✓ 2,000 AI requests
  • ✓ 1,000 min speech-to-text
  • ✓ 200 web searches
  • ✓ All 7 modes
  • ✓ Unlimited meeting debriefs
  • ✓ Priority support
Get Max
Ultra
$34/mo
For teams
  • ✓ 3,000 AI requests
  • ✓ 2,000 min speech-to-text
  • ✓ 300 web searches
  • ✓ All 7 modes
  • ✓ Unlimited meeting debriefs
  • ✓ Priority support
Get Ultra
Yearly License

All features, annual renewal. Resume intelligence, JD analysis, negotiation coaching, post-interview debriefs, 7 expert modes.

Get Yearly
Lifetime License ✦

One-time purchase. All future updates included. Own it forever.

Get Lifetime
Free to try · Windows · No card required
Download for Windows