Know if a voice is real
A cloned voice can authorize a wire transfer or pass a bank verification call. ORAVYS Anti-Deepfake screens recorded calls and audio files for synthetic voices and returns a documented result for each one.
- Documented results
- Web upload or API
- Audio analysed in memory on the site
3
Result bands
Plus a synthetic-voice score when conclusive
API
Recorded audio
One result per file
50 MB
Max upload
WAV, MP3, FLAC, OGG, M4A, WebM
$0
To start
1 free quick check per day
Think a voice is fake? Check it now.
Upload a clip and get a quick screening result. Free, no account needed.
The threat
Cloned voices now pass for real people.
A few seconds of recorded speech and a cheap online tool are enough to clone a voice. Banks, insurers, and call centers built their trust in voice before this was possible. Three kinds of fraud come up most often.
CEO and CFO fraud
A cloned executive voice, on a voicemail or a live call, asks finance staff to wire funds to a new account. In one widely reported case, a cloned executive voice obtained a transfer of 243,000 dollars.
Source: Wall Street Journal, 2019
Identity theft and voice KYC
A cloned customer voice can pass voiceprint checks in banking, telecom, and public-service call centers. A voice alone is no longer a safe way to prove who is calling.
Insurance and claims fraud
Synthetic recordings are sent as evidence with claims, accident reports, and compliance interviews. They sound clean, and their file details look ordinary.
Who uses it
One service, four kinds of work.
Each team sends the recordings it already has, in the browser or through the API, and gets one result per recording.
Banking and fintech
Voice approvals and call recordings
Check a recorded voice approval before acting on it, for example before a wire transfer. Your fraud team reviews the result alongside its other controls.
- Screening of voice approvals before payment
- One result per recording, by API
- Report page kept on file
Insurance and claims
Audio evidence
Check audio submitted with a claim before it is accepted as evidence. Each file gets a result and a report page your adjusters can attach to the case.
- Synthetic-voice screening per file
- Report page with file fingerprint and timestamp
- Web upload or API
Enterprise call centers
Recorded calls
Send recorded calls to the ORAVYS API and get a result per recording for your fraud team to review.
- Per-recording screening via API
- Recorded call files: WAV, MP3, M4A, WebM
- JSON results for your own tools
Media and newsrooms
Before publication
Check leaked audio, viral clips, and source recordings before you publish them, so a fabricated recording does not become your story.
- One result per clip
- Web upload for editors, API for desks
- Report page to keep with your notes
How it works
Upload, screen, read the result.
Each recording goes through several independent checks. No single check decides the result on its own.
01
Upload
Send a recorded file from the browser or through the API. WAV, MP3, FLAC, OGG, M4A, or WebM, up to 50 MB.
02
Screen
The recording is checked for signs of a synthetic voice by several independent checks.
03
Read the result
One result in three bands, with a synthetic-voice score when it is conclusive. JSON for the API, a report page for people.
Compliance
Documented for compliance teams.
Article 50 of the EU AI Act applies from August 2026 and sets transparency duties for synthetic audio. ORAVYS gives you a documented result for each recording you check.
EU AI Act
A written result per recording, to support your Article 50 transparency work.
GDPR
Data minimisation: on the site, audio is analysed in memory and not written to storage.
Retention
API jobs keep the file only until the result is written. Reports are deleted automatically. Details in the privacy policy.
Records
Each result has a report page with file fingerprint and timestamp, which your team can keep on file.
Pricing
Start free, then pay for what you check.
One free quick check per day. After that, pay per scan, buy a pack, or connect the API. No card for the free check.
Free
$0
1 quick check per day
- Quick screening, not a verdict
- Clips up to 30 seconds
- No card, no account
Single scan and packs
$9.90/scan
or $39.90 for a 5-scan pack
- Result in three bands, plus a synthetic-voice score when conclusive
- Report page with file fingerprint and timestamp
- No subscription
API plan
$99/mo
100 scans, 1,000 API calls a month
- Same result as a single scan, in your own software
- Sandbox keys self-serve; production keys on request
- Email support
Business and Enterprise: Business $899/month (2,000 scans, API and webhooks); Enterprise from $2,900/month, by contract, for banks, insurers, and governments. Current pricing
FAQ
Common questions.
What prospects and security teams ask us most.
How does the screening work?
Each recording goes through several independent checks, and no single check decides alone. You get one result in three bands, with a synthetic-voice score when it is conclusive.
What audio formats are supported?
WAV, MP3, FLAC, OGG, M4A, and WebM, up to 50 MB per file. A few seconds of clear speech are needed for a conclusive result; the free check accepts clips up to 30 seconds.
Is my audio stored?
On the site, audio is analysed in memory and not written to storage. API jobs keep the file only until the result is written. Reports are deleted automatically. The privacy policy lists every case.
Can it catch modern voice cloning?
It screens for voices produced by current text-to-speech and voice cloning tools. Use each result as one signal in your fraud process, next to your other controls.
Are the results admissible in court?
An automated result is a probabilistic signal, not a certification of authenticity, and does not decide admissibility on its own. Court admissibility is determined case by case by the court of competent jurisdiction (Daubert in US federal courts, Frye in some states, equivalent rules elsewhere) and depends on expert-witness qualification. For matters that need a signed, expert-reviewed opinion, see forensic examination on the pricing page. See our legal notices for the full disclaimer.
How fast is the API?
A short clip usually comes back while you wait; longer recordings take proportionally longer. Enterprise contracts set latency targets in the SLA.
How do I connect the API?
Send a recording to the REST API and receive a JSON result. Sandbox keys are self-serve, production keys are issued on request, and a Python SDK is available.
What about the EU AI Act?
Article 50 applies from August 2026 and sets transparency duties for synthetic audio. ORAVYS gives you a documented result for each recording you check, which you can keep with your compliance records.
Try it
Check the next suspicious call before you act on it.
Upload a recording now. One free quick check per day, no card, no account.
On oravys.com, audio is analysed in memory and not written to storage; API jobs and the exceptions are listed in the privacy policy. Supports deployer disclosure under EU AI Act Article 50(4).