Resemble AI

What Resemble AI does

Resemble AI is a generative AI security platform that protects enterprise decision-making by verifying, detecting, and acting on synthetic audio, image, or video content. The platform offers multimodal deepfake detection and watermarking solutions, available both on-premise and in the cloud. Resemble AI built its foundational voice AI models with provenance and detection in mind, giving it an advantage in deepfake detection. The company provides end-to-end generative AI security, from content creation through distribution, across various modalities. Its offerings include real-time detection, imperceptible watermarking, and biometric identity verification, all built on proprietary models and research. Resemble AI is committed to ethical AI use, with a robust consent system for voice cloning and strict prohibitions against harmful uses of AI voices.

Products

• Resemble DetectA 3 billion parameter multimodal detection model that covers audio, image, and video. It offers zero-day coverage for new generative models and provides explainable detection with human-readable forensic explanations.
• Resemble WatermarkerA multimodal, imperceptible watermarking solution for audio, image, video, and text. It embeds proof of ownership, authenticity, and verifiable provenance, supporting EU AI Act Article 50 compliance.
• Resemble IdentityA biometric speaker verification system that enrolls speakers from 4 seconds of audio and verifies or flags them in real-time across calls, recordings, or audio submissions to fight impersonation.
• Resemble MeetingsProvides real-time deepfake detection for virtual meetings on platforms like Zoom, Teams, Meet, and Webex, offering alerts and forensic audit trails.
• Resemble IntelligenceGenerates human-readable forensic explanations in real-time, surfacing artifacts that contribute to detection scores and providing context for compliance teams, legal review, and trust & safety workflows.
• ChatterboxAn open-source, production-grade Text-to-Speech (TTS) model with zero-shot voice cloning and emotion exaggeration control, embedded with PerTh watermarking by default.
• PerThAn open-source deep neural network that embeds imperceptible, psychoacoustically-masked watermarks into audio at the point of creation, surviving compression and editing.
• PerTh MultimodalAn enterprise and compliance-oriented multimodal model extending PerTh's watermarking across audio, video, image, and text in a single API, supporting real-time encode and decode.
• ResemblyzerAn open-source 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.

Key capabilities

• Multimodal deepfake detection (audio, image, video)
• Multimodal watermarking (audio, image, video, text)
• Real-time detection (<300ms)
• Zero-day model coverage for new generative AI models
• Explainable AI for detection results (human-readable reports)
• Biometric speaker verification and identity enrollment
• On-premise and cloud deployment options
• API access and SDKs
• Compliance with EU AI Act, GDPR, HIPAA, SOC 2 Type II, C2PA Standard
• Battle-tested against 160+ generative AI models
• Support for 51 languages
• Content provenance and traceability
• Voice cloning with consent mechanisms
• Integration with communication and collaboration platforms (Zoom, Teams, Meet, Webex)
• Web source verification for images
• Authenticity scoring and audit trails

Use cases

• Verify authenticity and protect IP across audio, video, and images
• Detect deepfakes in real-time to prevent damage
• Secure AI agents and voice agents
• Authenticate voice agents for telecommunications
• Prevent executive impersonation and wire fraud
• Detect deepfake interview and HR fraud
• Combat vishing and voice fraud at scale
• Spot and stop fraudulent claims in delivery, insurance, and financial services
• Ensure document integrity and compliance
• Protect talent voice, likeness, and productions in studios
• Confirm AI origin of content on platforms
• Achieve EU AI Act Article 50 compliance
• IP protection and rights management for media
• Content integrity and trust & safety for user-generated content
• Brand safety for spokesperson voice, ad creative, and campaign assets
• Remote onboarding identity verification

AI approach

Resemble AI builds foundational models for voice generation and deepfake detection across audio, video, and image. They develop proprietary models like DETECT-3B Omni for multimodal deepfake detection and PerTh for watermarking, and do not fine-tune existing models. They also offer open-source models like Chatterbox and Resemblyzer.

Tech named: DETECT-3B Omni, PerTh, PerTh Multimodal, Resemble Intelligence, Chatterbox, Resemblyzer, Gemini 3 Flash, Wav2Vec2, DALL-E 3, Midjourney, Stable Diffusion, Nano Banana, Flux, Black Forest Labs v2, Gemini, Imagen 3, GPT-4o, StyleGAN, Ideogram, Leonardo AI, Sora, Veo, Runway Gen-3 Alpha, HeyGen, Pika, Kling, Kuaishou, Seedance, ByteDance, Synthesia, D-ID, Bark, VALL-E, VALL-E2, YourTTS, Tortoise TTS, XTTS, StyleTTS 2, MetaVoice, Kokoro, VoiceBox, NaturalSpeech 3, HierSpeech++, Stable Diffusion XL, Latent Diffusion, eDiff-I, GLIDE, Kandinsky, DeepFloy

Industries served

• Telecommunications
• Banking & Finance
• Legal & Judicial
• Media & Entertainment
• Healthcare & HealthTech
• Corporate Security
• Delivery & Marketplace
• Insurance & Property
• Government

What it says sets it apart

• Only platform that verifies and detects across voice, image, and video for total security
• Built foundational voice AI models, providing an advantage in deepfake detection
• Proprietary models built from scratch, not fine-tuning existing ones
• DETECT-3B Omni is #1 on independent Podonos benchmark for audio deepfake detection
• PerTh watermarking is imperceptible, indestructible, and survives compression
• Offers explainable detection with human-readable forensic explanations (Resemble Intelligence)
• Open-source commitment: major capabilities have publicly available counterparts (Chatterbox, PerTh, Resemblyzer)
• Real-time detection in under 300ms
• Comprehensive compliance: EU AI Act, GDPR, HIPAA, SOC 2 Type II, C2PA Standard
• Supports air-gapped deployment for maximum security
• Ethical AI approach rooted in consent for voice cloning and strict misuse prohibitions

Funding rounds we track

From the AI funding tracker — rounds as reported by the linked publications.