CORE APPROACH
Physics
Passive sonar roots. No model to retrain.
Acoustic intelligence · on your desktop
Audar.
One window. Every sound.
Instant AI detection.
Detect synthetic content from ElevenLabs, Veo 3, Azure, Seedance and more — second by second. Physics, not prediction: the same audio always returns the same answer.
FREE CHECK · NO SIGN-UPONLINE CHECK
NO FILE HANDY? TRY A SAMPLE (WON’T USE UP YOUR FREE CHECKS)
- – synthetic
- – real
- – no speech
Your Audar.Net access key
You need this key to use your minutes in another browser or on another device. Keep it somewhere safe — it is not shown on your receipt.
Your file is analyzed and then deleted. Privacy Notice · Already bought minutes? Enter your access key
Find your access key
An assessment, not proof. Weigh it with source and context.
OUTPUT
Deterministic
Same audio, same answer. Every time.
PRIVACY
On-device
Desktop app: no voiceprints, no audio leaves the machine.
LANGUAGE
Any
Reads the signal, not the words or the speaker.
The problem
Voice cloning is no longer a research demo.
It takes seconds of audio to clone a voice, and the fakes arrive where you work: calls, meetings, voicemail, social video. Audar tells you what your machine is actually hearing.
01 —
The fraud is already here
Deepfake voice attacks target finance teams, executives, and families. The person on the call sounds exactly right — that's the point.
02 —
Your ear can't tell
Modern synthesis defeats human listening. Detection needs to read the physical structure of the signal, not the vibe of the voice.
03 —
Physics, not prediction
No machine learning, no model to retrain, no probabilistic guess — the same audio always returns the same answer.
04 —
Private by structure
Everything runs on your device. No voiceprints, no biometrics, no audio ever leaves the machine.
The approach
Why physics wins where AI falls short.
Detectors that chase generative models with more AI are locked in an arms race. Audar reads acoustic properties synthetic speech can't conceal — so it doesn't degrade as the generators improve.
Reading the ring
Synthetic voice — speech carrying indicators it may be AI-generated.
Real voice — speech consistent with a human voice.
No speech — silence, music, noise, or audio that can't be classified.
| ATTRIBUTE | AI VS AI | AUDAR · PHYSICS |
|---|---|---|
| Approach | Arms-race dynamics | Orthogonal |
| Detection method | Reactive, model-based | Structural |
| Answer | Probabilistic | Deterministic |
| Over time | Degrades | Stable |
| Compute | High | Low — runs on your desktop |
| Latency | High | Real-time |
Questions
Frequently asked questions.
Answers to common questions about Audar and about trying it free on this page.
Who is this page for?
This page gives anyone easy access to try Audar before subscribing. You can drop in an audio or video file and see how Audar works with three free checks, and no sign-up is required.
How does Audar work?
Audar uses acoustic analysis to examine the physical characteristics of sound. It looks for measurable properties and inconsistencies in the audio signal that help distinguish authentic human speech from synthetic or manipulated audio.
Does Audar work with any language or accent?
Yes. Because Audar analyzes the acoustic properties of sound rather than the words being spoken, it does not depend on language, accent, or any other biometric details of the speaker. Audar does not try to identify who is speaking; it only assesses whether the audio itself shows signs of being AI-generated.
What files can I check?
You can upload audio or video files of up to 75 MB, and Audar analyzes the first three minutes of each file. Supported formats are:
MP3 · WAV · AAC · M4A · OGG · FLAC · MP4 · MOV · WEBM · AVI · WMV · FLV · MPG
What has Audar been tested against?
Audar has been tested on AI-generated audio from ElevenLabs, Google Gemini, Microsoft Azure, Google Veo 3, and Seedance. It has also been tested on everyday content, including podcasts and YouTube videos.
What do the results mean?
Audar shows a ring divided into three colors, with the share of the file each one accounts for.
- Synthetic voice — speech carrying indicators that it may be AI-generated.
- Real voice — speech consistent with a human voice.
- No speech — silence, music, background noise, or audio that cannot be classified.
The large figure in the middle is the headline reading — Likely Synthetic or Likely Real — and it describes the speech in the file, not the whole recording.
Each result should be treated as an assessment rather than absolute proof that a recording is real or fraudulent. Audar's findings are most useful when considered alongside the source, context, and other available evidence.
What happens to the file I upload on this page?
Audio you upload on this page is discarded immediately after it is analyzed, and your results are not saved. Neither your audio nor your results are ever sent to third parties. For more detail, please see our Privacy Notice.
See it work
See the app before you buy it.
Audar scoring Admiral McRaven's Life Lesson and deepfake compilations, scored live as each plays. Third-party clips are used for illustration only; Audar is not affiliated with their creators.
Bad news · incident log
The attacks Audar exists for.
Pricing
One app. Two ways to pay.
MONTHLY
$149.95 /month
- — Full desktop application
- — Windows and macOS
- — Live session monitoring + file upload
- — Cancel anytime
No credit card needed
ANNUAL20% OFF MONTHLY
$1,440 /year
- — Everything in monthly
- — Priority support
- — $120/month, billed annually
No credit card needed
Try Audar free for 30 days, then subscribe from inside the app. Payments by Stripe. By subscribing you agree to our
Terms of Service and Privacy Notice.
Questions about volume or team licensing:
support@uncognito.ai
Early adopters
Want a human first?
We're onboarding early adopters directly. Write to us and someone from the team — not a form — sets you up.
support@uncognito.ai