Resemble AI

Resemble AI is a generative AI company that operates on both sides of synthetic media: it produces speech generation and voice cloning technology, and it builds models that detect AI-generated audio, video, images and text. Its text-to-speech and voice cloning products let users create a synthetic voice from a short recording, adjust pitch, speed, emphasis and emotion, and generate speech across a large number of languages, with output available as downloadable audio files or through an API for apps, games and websites.

The company's current commercial focus is its detection and verification platform, marketed as a way to secure enterprise generative AI from creation to distribution. Products include DETECT-World, a detection model for audio, video and images that combines pattern matching with a world model architecture; DETECT-3B Omni, an enterprise detection model the company reports at 98% accuracy across more than 38 languages; Resemble Intelligence, which produces human-readable explanations of detection scores and is powered by Google Gemini 3; Signal, which alerts users when content matches known fraud patterns without recording; Identity, for enrolling and matching voice and likeness; and PerTh Multimodal, an imperceptible watermarking system positioned for EU AI Act compliance.

Stated deployment scenarios include contact center and telephony fraud, real-time detection on meeting platforms, CSAM and sexualized deepfake triage, law enforcement forensics, KYC and account onboarding, document and receipt forgery, social platform trust and safety, and executive and brand impersonation. The company also publishes a database of reported deepfake incidents.

Business model

Resemble AI sells access to its generative voice and deepfake detection models to enterprises and developers, offering a free tier and self-serve sign-up alongside enterprise deployments. Delivery options span a single REST API covering all modalities, integrations with existing workflows and identity, SIEM and SOAR platforms, and fully containerized on-premises and air-gapped installations. Investors in its 2025 strategic round are also described as integrating the company's technology into their own platforms, creating distribution channels.

Sources describe paid access to Resemble AI's APIs and models, with a free entry tier and enterprise/on-premises deployments; specific pricing terms are not stated.

Traction

The company's application reports more than 4.5 million users. Its open-source Chatterbox repository has over 26,000 GitHub stars, Resemblyzer 3.3k and Resemble Enhance 2.4k; on Replicate its hosted models have accumulated large run counts, including 713.5K for Chatterbox Turbo, 621.4K for Resemble Enhance, 564.2K for its watermark model and 429.7K for Chatterbox. Customers are described as including global entertainment companies, Fortune 500 telecommunications providers and government agencies.

Latest developments

In December 2025 the company announced a $13 million strategic funding round, bringing total venture funding to $25 million, with investors including Google's AI Futures Fund, Sony Innovation Fund, Okta Ventures, Taiwania Capital, Gentree Fund, IAG Capital Partners, Berkeley Frontier Fund, KDDI Open Innovation Fund, Comcast Ventures, Craft Ventures, Javelin Venture Partners and Ubiquity Ventures. It subsequently released DETECT-World, its world-model-based detection system, and in August 2026 made the model available through Eden AI's unified API. It also published a set of 2026 deepfake threat predictions covering video-conferencing detection mandates, AI regulation readiness, identity security and deepfake insurance premiums.