Detect AI-generated and cloned voices at scale. One REST call returns a verdict, Authentic, Likely Synthetic, or Inconclusive, with a confidence-scored TrustScore, in seconds.
High accuracy · 60s file purge · Keys hashed at rest
A single POST to /v1/detect/voice covers AI-generated speech and voice clones alike. It is the single-modality companion to our full deepfake detection API, and powers the same engine behind our consumer AI voice detector.
Flags AI-generated speech and voice clones from across major voice engines, including ElevenLabs, PlayHT, Resemble, OpenAI, Murf, and WellSaid. The API detects synthetic origin without attributing audio to any specific engine.
Catches cloned voices and spliced or edited audio as well as fully synthetic speech, not just voices that are AI from scratch.
Each call returns one of three verdicts, Authentic, Likely Synthetic, or Inconclusive, alongside a TrustScore from 0 to 100 you can threshold on.
Pass a public or signed audio_url, or upload via multipart. We fetch, scan, score, and purge, all within 60 seconds of the verdict unless you opt into retention. Supported formats are MP3, WAV, M4A, FLAC, and OGG.
| Parameter | Type | Description |
|---|---|---|
| audio_url | string | HTTPS URL to the audio. Either audio_url OR multipart upload required.required* |
| file | multipart | Upload the audio directly as multipart form data. MP3, WAV, M4A, FLAC, or OGG.required* |
| webhook_url | string | Webhook for async jobs. We POST the verdict here when analysis is done. |
| strict_mode | boolean | Lower confidence threshold. Increases recall, may reduce precision. Default false. |
| retain | boolean | Keep the audio for audit instead of purging. Default false (60s purge). |
Verdicts map to Authentic / Likely Synthetic / Inconclusive. Files are deleted within 60 seconds of analysis unless retain is set.
Drop voice detection into your pipeline with batch processing, webhooks for async jobs, and a real-time streaming option for live calls.
Block vishing and voice-clone scams by flagging synthetic callers in real time across your call center.
Screen synthetic and cloned voices during voice onboarding and identity verification flows.
Verify audio and interview recordings before publication and triage suspect clips at speed.
Check meeting recordings and voicemail for cloned voices before they trigger a payout or approval.
Process audio one clip at a time or in batch, register a webhook_url for async results, and use the real-time streaming option for live call and contact-center scenarios. Want the how-to? Read how to detect AI voices and our guide to voice cloning scams.
Same call to detect an AI-generated voice, same response shape. Auth, retries, exponential backoff, file streaming, and webhook signature verification handled for you.
Start free with 50 detections per month, no card required. API access begins on Starter at $49/mo, Business is $199/mo, and Enterprise is $599/mo. See full limits on our pricing page.
Free. 50 detections/month in the web app, no card required.
Starter. Adds API access for programmatic voice detection.
Business. Higher volume, higher rate limits, batch workflows.
Enterprise. Highest throughput with priority support.
Everything teams ask before they integrate. New to this? Learn how to detect AI voices.
A REST API that analyzes audio and returns whether a voice is AI-generated or cloned, or authentic, with a confidence-scored verdict.
Synthetic speech from major engines including ElevenLabs, PlayHT, Resemble, OpenAI, Murf, and WellSaid.
POST an audio_url or upload a file to /v1/detect/voice with your API key; you get back a JSON verdict (Authentic / Likely Synthetic / Inconclusive) and a TrustScore.
MP3, WAV, M4A, FLAC, and OGG; limits depend on your plan.
Yes, use streaming and webhooks for live call and contact-center scenarios.
Yes, teams use it to flag synthetic and cloned voices in calls, onboarding, and voicemail as part of a fraud and trust pipeline.
A free account includes 50 detections/month; API access starts on the Starter plan. Audio is deleted from primary storage within 60 seconds of analysis unless you opt into retention.
One REST endpoint, confidence-scored verdicts, webhooks, and SDKs, built for fraud and trust & safety teams.