oravys 0.1.0

Voice deepfake detection SDK - analyze any audio in 5 lines of Python. 3000+ forensic engines. By Oravys Inc.

ORAVYS - Voice Deepfake Detection SDK

Detect synthetic and cloned voices in 5 lines of Python. 3000+ forensic engines.

Install

pip install oravys

For local analysis with advanced DSP:

pip install oravys[local]

Quick Start

API Mode (3000+ engines)

from oravys import Detector

det = Detector(api_key="sk-oravys-...")
result = det.analyze("call_recording.wav")

print(result.verdict)        # "SYNTHETIC" or "AUTHENTIC"
print(result.confidence)     # 0.97
print(result.engines_fired)  # 847

Local Mode (offline, ~11 DSP engines)

from oravys import Detector

det = Detector(mode="local")
result = det.analyze("voice.wav")

print(result.is_synthetic)   # True/False
print(result.confidence_pct) # "72.7%"

CLI

# Local analysis
oravys detect voice.wav --verbose

# API analysis
oravys detect voice.wav --api-key sk-oravys-... --json

# Version
oravys version

Local Engines

The local mode runs these lightweight DSP engines (CPU-only, no GPU needed):

Engine	Domain	What it detects
silence_ratio	temporal	Unnatural silence patterns
zero_crossing_rate	temporal	Synthetic signal characteristics
spectral_flatness	spectral	Noise-like spectrum (codec artifacts)
pitch_stability	prosodic	Unnaturally stable pitch (TTS signature)
harmonic_noise_ratio	spectral	Overly clean harmonics
spectral_bandwidth	spectral	Bandwidth consistency anomalies
energy_contour	temporal	Robotic energy patterns
formant_transition	articulatory	Missing natural articulation
micro_pause	prosodic	Absence of natural hesitations
breathiness	phonatory	Missing breath variation
tremor	neuromotor	Missing natural vocal tremor

The full API mode runs 3000+ engines across 16 forensic domains.

Get an API Key

Sign up at app.oravys.com to get your API key.

License

Proprietary. Copyright (c) 2026 Oravys Inc. All rights reserved.