Resemble AI

Resemble AI: The Trust Layer for Voice AI

Resemble AI provides deepfake detection, watermarking, and identity verification for teams building with voice AI. Founded in 2019 as a voice generation company, Resemble spent years learning how synthetic speech is made, and now applies that expertise entirely to detection. Its APIs let voice agents verify who (or what) is on the other end of a call, mark AI-generated audio at the source, and help voice products meet emerging transparency and compliance requirements.

Key Features

• Real-time deepfake detection: Detect synthetic speech in live calls and recordings with DETECT-World, benchmarked against 250+ generators, including zero-day coverage of new ones.

• Agent detection: Determine whether your voice agent is talking to a human or another AI agent before spending time and tokens on the conversation.

• Watermarking at the source: Embed imperceptible PerTh Multimodal watermarks into AI-generated audio, with C2PA support for verifiable provenance.

• Identity verification: Enroll voices with consent and authenticate speakers against them.

• Explainable results: Intelligence pairs every detection with a plain-language explanation your team or system can act on.

• Fraud signals: Match calls against known fraud patterns with Signal.

• Built for compliance: Support transparency requirements under the EU AI Act and California's SB 942. Resemble is a signatory to the EU's Code of Practice on Transparency of AI-Generated Content.

• Developer-first: API-first architecture and SDKs, with cloud or on-prem deployment.

Use Cases

• Voice agents that screen inbound callers for synthetic voices and other AI agents

• Contact centers defending against voice-clone fraud and account takeover

• Voice generation platforms watermarking output for provenance and compliance

• Authentication flows with consent-based voice verification

• Trust and safety teams flagging synthetic audio in user uploads

• Meeting and communication tools guarding against executive impersonation

Model Selection

• DETECT-World: Multimodal deepfake detection across audio, video, and image

• PerTh Multimodal: Watermarking and C2PA provenance

• Intelligence: Explainable analysis of detection results

• Identity: Voice and likeness enrollment and authentication

• Signal: Fraud pattern recognition and matching

• Agent Detection: LLM, crawler, and agent identification

Getting Started

• Website: resemble.ai

• API & SDK: Developer Docs

• Builder's Grant: Up to $30,000 in credits

• Contact & Demo: Contact Resemble AI

Resemble AI helps voice AI teams build products people can trust, with detection, provenance, and identity built in from the first call.