Voice clone and synthetic speech detection

Cloned voices are now used in fraud calls, fake statements and social engineering. sealverity.ai analyses recorded audio and live call audio for signs of synthetic speech, using several detectors and showing scores segment by segment so analysts can see exactly which parts of a recording are suspicious.

Segment-by-segment audio scoring

Uploaded audio is split into segments and each segment is scored. The result is shown on a waveform with per-segment scores, so a recording that is mostly genuine but contains an inserted synthetic sentence is not hidden behind an average.

As with images and video, audio is sent to every enabled detector in parallel and combined with audio-specific weights. When detectors disagree the result is marked inconclusive and routed to review.

Call-centre voice check API

Contact centres and fraud teams can send short audio chunks to the voice check API and receive a result usually within a few seconds. Chunks of two to ten seconds work best. Detectors slower than the time limit are skipped for that chunk so the caller is not kept waiting, and minutes analysed are counted for billing.

The API uses organisation API keys created in settings, and full request and response details are in the developer documentation.

REST endpoint for short audio chunks
Results usually returned in a few seconds
Per-detector scores and the combined score in the response
Same weights, thresholds and audit trail as the web app

Live calls

For meetings, sealverity.ai can place a bot in Zoom, Microsoft Teams or Google Meet that scores each participant's audio every 10 seconds. Some detectors need longer audio, so they receive a rolling 20 to 30 second window while the on-screen gauge still updates every 10 seconds. Live calls have their own threshold and their own detector weights, because compressed meeting audio behaves differently from studio recordings.

Evidence you can stand behind

Every audio file is hashed at intake and recorded in a hash-chained chain of custody. Analysts record their decision and reason, and the result can be included in a signed forensic report that anyone can verify publicly. Scores support an analyst's judgement; they are never presented as a verdict on their own.

Who uses it

Investigators checking leaked or viral recordings, compliance teams reviewing payment-authorisation calls, newsrooms verifying audio before publication, and contact centres screening callers who may be using a cloned voice of a customer or executive.

Part of one investigation platform

This capability is part of sealverity.ai, a single platform for media authenticity and investigations. Findings from images, video, audio, live calls, identity checks and social and news monitoring can be linked to cases, entities and a link graph, with risk scores that show which factors contributed. Access is protected by mandatory multi-factor authentication, single sign-on and SCIM provisioning, five roles, retention policies and legal hold. Organisations can bring their own self-hosted detectors or switch on sovereign mode, which blocks every external call on the server and logs each blocked attempt. Integrations include a browser extension, Slack and Teams, email scanning, webhooks, SIEM connectors and a REST API.