For Developers

Detect AI-generated voice in 300 ms. Drop-in REST API.

Send an audio file, get back an authenticity verdict plus per-engine risk scores. 3000+ forensic engines run in parallel, fused into a single calibrated score. Python / Node / Go SDKs available.

Get an API key Talk to engineering
# Python SDK example
from oravys import Client

client = Client(api_key="sk_...")
result = client.audio.analyze(file="call_recording.wav")

print(result.is_authentic)       # True / False
print(result.confidence)         # 0.0 .. 1.0
print(result.top_engines)        # list of contributing detectors
# cURL one-liner
curl -X POST https://app.oravys.com/api/sdk/analyze \
  -H "X-API-Key: sk_..." \
  -F "audio=@call.wav"

300ms median verdict

Streaming response, keep-alive bytes for long files. Sub-3s p99 even at 5-min audio.

3000+ engines

Multi-engine ensemble: DSP + SSL + LLM judges. Disagreement-aware fusion.

SDK clients

Python, Node, Go. Postman collection. OpenAPI 3.1 spec. SDK error contracts.

Free tier

20 calls/month included on Pro $19. Higher quotas on Max ($99) and Max+ ($249). Custom on Enterprise.

Open source companion: VoiceSign

Watermark your own voice recordings before publishing them. The mark is imperceptible to humans but survives compression and is verifiable by any ORAVYS detector.

pip install voicesign

MIT-licensed Python library. Drop-in, no API key required for client-side watermarking. Verification still hits our API.