Free Tier for Researchers

ORAVYS
ACADEMIC API

107 analysis engines for voice & speech research. Built for phonetics, prosody, and computational paralinguistics. REST API with JSON output, batch support, and full reproducibility.

Interspeech-Ready IEEE-Cited ICASSP-Compatible Open Science
107 Engines
137 Acoustic Features
1.1M Parameters
99.96% F1 Score
Live Analysis 137 Features
POST https://api.oravys.com/v1/analyze

Simple by Design

One endpoint. One audio file. 107 engines run in parallel. Results in JSON with sub-second latency on standard research hardware.

Shell / cURL
curl -X POST https://api.oravys.com/v1/analyze \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: multipart/form-data" \
  -F "audio=@sample.wav" \
  -F "engines=all" \
  -F "format=json"
JSON Response
{
  "status":             "success",
  "engine_count":       107,
  "feature_count":      137,
  "duration_ms":        843,
  "authenticity_score": 0.9312,
  "emotion_profile": {
    "dominant":  "neutral",
    "valence":   0.62,
    "arousal":   0.31,
    "confidence": 0.88
  },
  "features": {
    "f0_mean":       182.4,   // Hz
    "f0_std":        28.7,    // Hz
    "jitter_local":  0.0041,  // ratio
    "shimmer_local": 0.071,   // dB
    "hnr":           18.3,    // dB
    "cpp":           12.1,    // dB
    "mfcc":          [-312.4, 82.1, -14.2, "..."],
    "formants":      { "F1": 731, "F2": 1182, "F3": 2638 }
  },
  "engines": {
    "DeepfakeDetector":  { "score": 0.04, "label": "genuine" },
    "EmotionEngine":     { "dominant": "neutral", "valence": 0.62 },
    "PersonalityEngine": { "openness": 0.71, "conscientiousness": 0.64 },
    "...": "104 more engines"
  },
  "model": {
    "version":    "V4",
    "f1_score":   0.9996,
    "parameters": "1.1M",
    "arch":       "BiLSTM+TransformerEncoder+AttentionPool"
  }
}

107 Engines. 9 Categories.

Each engine is independently registered via @register_engine, runs in parallel, and returns a typed JSON object. Hover any card to see engine names.

⚙️
Core 21

Foundational authenticity detection and deepfake classification. Primary BiLSTM+Transformer pipeline with GroupSplit certification.

DeepfakeDetector AuthenticityScorer VoiceFingerprint SpeakerVerifier SignalQuality NoiseAnalyzer +15 more
🔬
Forensic 28

Audio forensics, manipulation detection, compression artifact identification, and chain-of-custody integrity analysis.

CompressionArtifact SplicingDetector RoomAcoustics MicrophonePrint GlitchDetector +23 more
🩺
Clinical 15

Vocal biomarkers for health research. Stress physiology, dysphonia markers, and vocal fatigue indices. EU AI Act Article 50 compliant.

StressIndex DysphoniaMarker VocalFatigue BreathPattern VocalTremor +10 more
🧠
Personality 16

Big Five, HEXACO, and MBTI acoustic correlates. Prosody-to-trait mapping from peer-reviewed models.

BigFiveVoice HEXACOProfile MBTIAcoustic OpennessIndex +12 more
💼
Professional 5

Workplace communication quality, leadership acoustics, and persuasion dynamics for organizational research.

LeadershipVoice PersuasionScore CommunicationQuality AuthorityIndex +1 more
🎙️
Voice Quality 11

Acoustic quality metrics: HNR, jitter, shimmer, CPP, GNE. Full phonatory analysis compatible with Praat exports.

HNRAnalyzer CPPEstimator GNEComputer JitterShimmer +7 more
⏱️
Temporal 3

Time-series analysis of vocal changes: longitudinal tracking, spectral trajectory, and rhythm dynamics.

LongitudinalTracker SpectralTrajectory RhythmDynamics
🔗
Relationships 4

Dyadic interaction patterns, entrainment, and conversation dynamics for social speech research.

AcousticEntrainment DyadicDynamics TurnTaking EmotionalSync
137

Acoustic Features Extracted

Every analysis computes the full 137-dimensional feature vector. Grouped below by perceptual category with Praat-compatible naming.

Spectral
MFCC 1–20 MFCC Δ MFCC ΔΔ MelSpec STFT Spectral Centroid Spectral Flux Spectral Rolloff Spectral Contrast Chroma CQT Tonnetz
Prosodic
F0 Mean F0 Std F0 Range F0 Slope Intonation Contour Speaking Rate Pause Ratio Rhythm Regularity Stress Patterns Tempo
Voice Quality
Jitter (local) Jitter (rap) Shimmer (local) Shimmer (dB) HNR CPP GNE NHR SHR Vocal Fry Breathiness Creakiness
Formants & Articulation
F1 Mean/Std F2 Mean/Std F3 Mean/Std F4 Formant Ratios Vowel Space Articulation Rate VSA Index LPC Coefficients
Temporal / Energy
RMS Energy ZCR Energy Envelope Loudness (LUFS) Dynamic Range Onset Strength Beat Salience Silence Ratio
Deep Embeddings
BiLSTM Hidden Transformer CLS AttentionPool Vec wav2vec2 Feats x-vector d-vector

Built on Peer-Reviewed Science

Every engine references validated literature. Designed for reproducibility and publication-ready output.

Key References

[1] DePaulo et al. (2003). Cues to deception. Psychological Bulletin, 129(1), 74–118.
[2] Schuller et al. (2013). The INTERSPEECH 2013 Computational Paralinguistics Challenge. Proc. Interspeech.
[3] Eyben et al. (2016). The Geneva Minimalistic Acoustic Parameter Set (GeMAPS). IEEE TASLP, 24(1).
[4] Baird et al. (2020). The Interspeech 2020 Computational Paralinguistics Challenge. Proc. Interspeech.
[5] Vaswani et al. (2017). Attention is all you need. Advances in Neural Information Processing Systems, 30.
[6] Murray & Arnott (1993). Toward the simulation of emotion in synthetic speech. JASA, 93(2), 1097–1108.

Used in Research At

Interspeech ICASSP IEEE TASLP Speech Comm. Computer Speech & Lang. EACL

Cite ORAVYS

BibTeX
@software{oravys2026,
  author    = {ORAVYS Ltd.},
  title     = {ORAVYS: 107-Engine Voice Analysis API
             for Speech Research},
  year      = {2026},
  version   = {V4.0},
  url       = {https://app.oravys.com/academic},
  note      = {BiLSTM+TransformerEncoder, F1=0.9996,
             137 acoustic features, 9 domains}
}
⚖️
EU AI Act — Article 50 Compliant

ORAVYS is classified as a research tool, not a lie detector. All clinical outputs are labeled as biomarkers requiring qualified interpretation. Lawful use in EU academic contexts confirmed.

Accessible for Every Researcher

All tiers include full access to all 107 engines, JSON export, and batch processing. Upgrade or cancel at any time.

Free
$0
forever · research only
100 analyses / month
  • All 107 engines
  • 137 acoustic features
  • JSON export
  • Single file upload
  • Academic use license
Apply for Access
Institution
Custom
annual contract · multi-seat
Unlimited analyses
  • Everything in Academic
  • Unlimited batch jobs
  • On-premise deployment
  • Dedicated GPU allocation
  • SLA 99.9% uptime
  • Data Processing Agreement
  • Named researcher support
Contact Sales

Comprehensive Docs

OpenAPI 3.1 spec, interactive playground, and annotated examples for every endpoint.

Ready to start?

Get your free API key and run your first analysis in under 2 minutes.

Request Free API Key
GUEST