Complete Generative AI Security: Detect, Verify & Generate

GENERATE

Human-quality voice from ten seconds of audio

ORIGINAL

Angela Merkel, real

CLONE

Angela Merkel, cloned

Resemble Text-to-Speech

100 languages

Emotion and tone control

Sub-200ms latency

LEARN MORE

Resemble Voice Creation

Clone from 10s of audio

Voice design (prompt-to-voice)

Consent workflows

LEARN MORE

Resemble Audio

Edit audio by typing

Audio inpainting

Speech-to-speech

LEARN MORE

Resemble Speech-to-Speech

Speaker diarization

Sentiment analysis

Language detection

LEARN MORE

VERIFY

Protect your content before it leaves your hands.

Prove authenticity and protect IP across audio, video, and images with imperceptible watermarks that survive compression, and biometric identity verification that works in real time.

Resemble Identity

Biometric speaker verification

Enroll from 4 seconds

Cross-app auth

LEARN MORE

Resemble Watermarker

Audio, video, and image

EU AI Act Art. 50 compliant

Inspector SDK

LEARN MORE

DETECT

Stop deepfakes before they cause damage.

Real-time multimodal detection across audio, video, and images. Battle-tested against 160+ generative AI models with explainability built in so your compliance team can trust every verdict.

Resemble Detect

Audio, video, and image

96.7% detection accuracy

Tested against 160+ Gen-AI models

LEARN MORE

Resemble Meetings

Zoom, Teams, Meet, Webex

Real-time alerts

Forensic audit trail

LEARN MORE

FOUNDATIONAL MODELS

Built on original research

We don't fine-tune existing models, we build from scratch. Every product is powered by a proprietary model trained on our own data and research.

Chatterbox Turbo

GENERATE

350M parameter architecture optimized for voice agents. 1-step decoder, native paralinguistic tags — [cough], [laugh], [chuckle]. Built for latency-critical production.

Real-time

350M params

Paralinguistic

ONNX available

GET STARTED GITHUB HUGGING FACE

PerTh Watermarker

VERIFY

Perceptual Threshold — a deep neural network that embeds imperceptible, psychoacoustically-masked watermarks into audio at generation time. Survives MP3 compression, audio editing, noise, and codec transforms. Embedded in every Chatterbox output upon request.

~100% detection accuracy

Survives compression

Audio

On-prem

GET started DOCS

DETECT-3B Omni

DETECT

3 billion parameter multimodal detection model. The only deepfake detector covering audio, image, and video in a single unified architecture. Zero-day coverage for new generative models in under an hour.

#1 benchmarked detection

<300ms detection

3B parameters

160+ models

51 languages

On-prem

GET STARTED API Docs Research BENCHMARKS

Resemble Intelligence

DETECT

Intelligence generates human-readable forensic explanations in real time, surfacing which artifacts contribute to scores and why. Full explainability, context and Reverse Search provide the most up-to-date and complete rationale for results. Built for compliance teams, legal review, and trust & safety workflows that need evidence, not just a score.

Human-readable reports

Real-time

Audit-ready

Docs

OPEN SOURCE

If we built it, we ship the tools to catch it.

Our open-source commitment: every major capability we release has a publicly available counterpart. Freely licensed, production-ready, and actively maintained.

Chatterbox

MIT OPEN SOURCE

Production-grade TTS with zero-shot voice cloning. Outperforms ElevenLabs in blind evaluations. First open-source model with emotion exaggeration control.

Sub-200ms

Voice cloning

Emotion control

PerTh watermarker available

Get started Docs Github HUGGING FACE

Resemblyzer

APACHE 2.0

Deep learning voice encoder that derives a high-dimensional speaker representation from a few seconds of audio. Used for voice authentication, speaker diarization, and similarity scoring.

Speaker embedding

Voice similarity

Diarization

Github

INTEGRATIONS AND API

Fits into what you already have

Full REST API, official SDKs, and direct integrations. Deploy in the cloud, on-prem, or fully air-gapped. Every pillar accessible through a single unified API.

View all integrations Full API docs

AWS

EC2, Lambda, SageMaker

Azure

AKS, Azure ML

Google Cloud

GKE Vertex AI

Docker / K8s

Self-hosted containers

Air-gapped

No external calls

SSO / SAML

Okta, Azure AD

Webhooks

Event-driven callbacks

Streaming API

Real-time TTS and detect

SECURITY AND COMPLIANCE

Enterprise-grade. Out of the box.

Built for environments where security is non-negotiable — from air-gapped infrastructure to international data regulations.

SOC 2 Type II

Independently audited security controls covering availability, confidentiality, and data integrity.

In-progress

GDPR

Fully compliant with EU data protection regulations. No personal data processed without lawful basis.

Compliant

HIPAA

Supports healthcare and public health agency deployments with full HIPAA-aligned data handling.

Compliant

ISO 27001

Internationally recognised standard for information security management systems.

In-progress

Air-gapped deployment

Fully containerized on-premises install. No external connections. No data leaves your environment ever.

Available

Deploy in 24 hours

Guided installation wizard gets you from contract to live detection in under a day — not months.

On-prem & cloud

Zero retention mode

Submitted media is permanently deleted after detection completes — no retention, no re-analysis. Meets the strictest data sovereignty requirements.

Available

API-first architecture

Single REST API covering all modalities. Integrates with existing SIEM, SOAR, and identity platforms.

Production-ready

Frequently asked questions

How does Resemble AI detect deepfakes?

DETECT-3B Omni is a 3-billion parameter model covering audio, image, and video in a single architecture. It returns a verdict in under 300ms, tested against 160+ generative AI models, and ranked #1 on the Speech DeepFake Arena benchmark. Resemble Intelligence pairs that output with a human-readable forensic explanation, identifying which artifacts triggered the flag and why.

What is the difference between Resemble Detect and Resemble Meetings?

Detect processes submitted audio, video, and image files via API and returns a detection verdict with forensic detail. Resemble Meetings integrates directly with Zoom, Microsoft Teams, Google Meet, and Webex to flag synthetic voice in live calls, with real-time alerts and a per-session audit trail. Detect is for asynchronous file review. Meetings is for live call protection.

What is PerTh watermarking and how does it work?

PerTH embeds an imperceptible signal into AI-generated audio at the perceptual threshold. The watermark is constrained to speech-relevant frequencies, making removal without audible corruption extremely difficult. It survives resampling, re-encoding, compression, noise addition, pitch shifting, and time-stretching. Detection returns a binary result: watermark present or not present.

Can Resemble AI be deployed on-premise or in an air-gapped environment?

Yes. All products support on-premise deployment via Docker and Kubernetes. Air-gapped installations make no external connections.

What happens to my data when I use Resemble AI?

Submitted media is not stored or used for model training without explicit consent. Zero Retention Mode permanently deletes audio after analysis completes. Air-gapped deployment ensures no data leaves your environment. Resemble AI is GDPR compatible and HIPAA compatible.

Is Resemble AI compliant with the EU AI Act?

Resemble Watermarker aligns with EU AI Act Article 50, which requires watermarking on AI-generated content and takes effect August 1, 2026. The PerTH watermarker (audio) and Videoseal (image/video, MIT license) provide the persistent, machine-readable provenance signal the regulation requires.

End-to-end generative AI security

Protect your content before it leaves your hands.

Stop deepfakes before they cause damage.

Built on original research

If we built it, we ship the tools to catch it.

Fits into what you already have

Enterprise-grade. Out of the box.