
FARx delivers industry-leading accuracy across both voice and facial recognition modalities. Our voice biometric engine achieves a False Acceptance Rate (FAR) of less than 0.1% and a False Rejection Rate (FRR) below 1% under standard conditions. Our facial recognition engine achieves 99.7%+ accuracy across diverse demographics, lighting conditions, and device types. Both engines are continuously trained and validated against real-world datasets to maintain performance at scale.
Note: These accuracy rates are achieved once the system is fully tuned to your environment, user population, and deployment conditions. FARx’s implementation team works with each client during onboarding to optimise thresholds and ensure peak performance before go-live.
FARx does not apply noise suppression or image enhancement, as pre-processing can interfere with the vocal and facial characteristics the models depend on and produce worse outcomes than leaving the input untouched. Instead, FARx uses speech and image quality checks to assess each sample and passes the results to the calling application as part of the authentication response. The business process can then decide how to proceed — with full visibility that lower quality input may produce a less reliable result. The quality check pipeline also incorporates liveness detection, including identification of synthetic voice, so the application is informed of any spoofing signals alongside the quality assessment.
FARx’s approach to liveness detection goes beyond conventional anti-spoofing. For voice, FARx has been specifically trained on an extensive database of synthetic voices, enabling the models to distinguish between a real human voice and a cloned or AI-generated one. This training-based approach means the synthetic voice detection models can be continuously updated as new voice generation technologies emerge, ensuring FARx remains effective against the latest cloning and deepfake tools. For facial liveness, FARx counters photo, video, and 3D mask attacks using passive and active challenge-response techniques.
FARx is a text-independent voice biometric system — it verifies speakers from natural conversational speech with no requirement for a specific phrase or scripted input. Unlike speech recognition, which identifies what is being said regardless of speaker, voice biometrics identifies who is speaking regardless of what is said. To optimise accuracy, FARx models are tuned to the acoustic attributes of specific languages and have been deployed across a wide range of languages and regional accents in production environments worldwide.
Yes. FARx supports both modes. Speaker verification is a 1:1 match — confirming that a caller is who they claim to be against a single enrolled profile, used for authentication. Speaker identification is a 1:N match — comparing an incoming voice against a watchlist of known fraudsters to detect a match, even when the caller presents under a different identity. Both can be deployed independently or together within the FARx platform.
FARx supports both deployment models. Our cloud-hosted SaaS offering provides rapid deployment, automatic updates, and elastic scalability — ideal for organisations that want to go live quickly with minimal infrastructure overhead. For enterprises with strict biometric data sovereignty, regulatory, or security requirements, FARx is also available as a fully on-premise deployment that runs within your own data centres or private cloud environment. Hybrid configurations are also supported for organisations that require flexibility across multiple environments.
FARx is fully modular. Clients can deploy voice biometrics only, facial recognition only, or a combined multimodal solution. Multimodal deployment offers the highest level of assurance by requiring both modalities for authentication, which is particularly suited to high-risk or regulated use cases. Many clients begin with a single modality and add the second as their biometric programme matures. Licensing is structured to support each configuration independently.
Absolutely. The two technologies are complementary by design. Speech recognition identifies what is being said, irrespective of who is speaking. FARx identifies who is speaking, irrespective of what is being said. This fundamental difference means FARx can be layered directly on top of any existing speech recognition solution to add identity assurance and security without disrupting the underlying capability. Typical deployment scenarios include telephony and contact centre speech recognition services, dictation systems, financial trading platforms, and any application where speech recognition is already in use. In each case, FARx secures the session by continuously verifying the identity of the speaker in the background.
FARx uses a layered security architecture to protect all biometric data. Biometric templates are encrypted at rest using AES-256 and in transit using TLS 1.3. FARx employs cancelable biometrics — irreversible one-way transformations of raw biometric data — so that if a template is ever compromised, it can be revoked and re-enrolled without exposing the original biometric. All cryptographic keys are managed through FIPS 140-2 validated key management systems, and access to biometric data is strictly role-based and audited.
Yes. FARx integrates natively with MFA frameworks, allowing biometrics to serve as one factor alongside PINs, OTPs, hardware tokens, or device-based authentication. FARx can function as a standalone authenticator or as part of a broader identity assurance stack, including integration with existing IAM and SSO platforms. This flexibility allows enterprises to define the right level of assurance for each use case and user population.
FARx treats biometric data as a first-class sensitive asset. Raw biometric data (voice recordings and facial images) is processed transiently and is not stored by default — only the derived, encrypted template is retained. Clients retain full control over template storage policies, including the ability to store templates within their own environment rather than FARx's infrastructure. All processing is logged and auditable, and data minimisation principles are applied throughout the pipeline.
Yes. FARx is designed to support compliance with major biometric and data privacy regulations, including the EU General Data Protection Regulation (GDPR), the Illinois Biometric Information Privacy Act (BIPA), the California Consumer Privacy Act (CCPA), and equivalent frameworks in other jurisdictions. FARx provides data processing agreements (DPAs), records of processing activities (ROPAs), and consent management tooling to help clients meet their regulatory obligations. We recommend that clients work with their legal and compliance teams to ensure their specific deployment configuration aligns with applicable law.
Yes. FARx supports right-to-erasure workflows that allow individuals to have their biometric templates permanently deleted upon request. Deletion is cryptographically verified and logged, providing an auditable trail for compliance purposes. Clients can configure automated data retention and expiry policies to align with their internal governance frameworks and applicable regulatory requirements.
No. FARx does not sell, share, or monetise biometric data. Biometric templates and associated data belong exclusively to the client and are used solely to provide the contracted service. FARx operates as a data processor under applicable privacy frameworks, and all data handling obligations are defined in the Data Processing Agreement provided at contract execution.
FARx maintains a comprehensive compliance posture appropriate for enterprise and regulated industry deployments. Certifications and standards include ISO 27001 (Information Security Management), SOC 2 Type II, and FIPS 140-2 validated cryptographic modules. FARx also supports deployments within FedRAMP-aligned environments for US public sector clients. Our compliance documentation and audit reports are available to qualified prospects under NDA.
Yes. FARx is purpose-built for high-assurance environments. Our financial services clients use FARx for customer authentication, fraud prevention, and KYC/AML workflows. Healthcare deployments leverage FARx for workforce access control and patient identity verification. Government and public sector clients benefit from our on-premise deployment option and our alignment with NIST SP 800-63B digital identity guidelines. FARx's compliance framework is designed to adapt to the specific regulatory landscape of each client's sector.
FARx integrates via a REST API and web SDK, with native iOS and Android SDKs in development. IAM platform connectors — including Okta, Microsoft Entra ID, Ping Identity, and ForgeRock — and federated identity support via SAML 2.0 and OIDC are implemented through FARx’s system integration partners. Dedicated integration guides are available for the three most common deployment patterns: FARx for Websites, Portals & Web Applications; FARx for Telephony, IVR & Contact Centres; and FARx for Microsoft Teams — available on request from your FARx account representative.
Implementation timelines vary based on deployment model, integration complexity, and client environment. A cloud-based proof of concept can typically be stood up within one to two weeks. A production deployment with full API integration generally takes four to twelve weeks, depending on the scope of existing system integration. FARx’s deployment methodology goes beyond technical integration — it includes user enrolment and tuning of the FARx voice processing background models to optimise accuracy for that specific population of speakers. This population-level tuning is a critical step that ensures the system is not only integrated but secure and performing at its best before go-live.
Yes. FARx provides a fully featured sandbox environment for development, testing, and evaluation purposes. The sandbox mirrors the production API and includes synthetic test data sets for both voice and facial modalities. Access to the sandbox is available to qualified prospects and active clients, and is typically provisioned within 24 to 48 hours of request.
FARx is priced on a consumption basis, structured around enrolled users and monthly authentication volumes, with volume discounts and multi-year rates available. On-premise deployments are offered under perpetual licence or annual subscription. Unlike many competitors, deepfake voice detection is included as standard within the voice quality and liveness pipeline at no additional cost. Pricing is provided through a formal proposal following a scoping conversation with our enterprise sales team.
FARx enterprise agreements are available on annual terms, with multi-year options providing the most favourable commercial terms. We do not impose a minimum user count for standard enterprise agreements, though certain pricing tiers are unlocked at higher enrolment volumes. Proof of concept engagements are available under a separate short-term evaluation agreement for qualified prospects.
Yes. Each modality is licensed independently, and multimodal deployments are priced as a bundled offering that is more cost-effective than licensing each modality separately. Clients who later expand from a single modality to multimodal can do so under a modular upgrade path without requiring a new commercial agreement. Detailed pricing for each configuration is available upon request.
FARx and speech recognition solve different problems: speech recognition identifies what is being said; FARx identifies who is saying it. That makes them natural partners. FARx can be layered on top of any existing speech recognition deployment to add continuous identity assurance and synthetic voice protection. The ten most common use cases are below.
1. Contact Centres & IVR
Speech recognition handles routing and agent assist. FARx runs in the background verifying the caller’s identity throughout the call and screening against fraud watchlists — catching known fraudsters even when they use a different name. Synthetic voice detection identifies deepfake and cloned voices in real time, closing the gap that traditional authentication leaves wide open.
2. Dictation & Documentation
Dictation systems capture what is said; FARx confirms who said it. In medicine, law, and regulated business, records must be attributable to a named individual — FARx provides that assurance continuously. Synthetic voice detection ensures that AI-generated or cloned audio cannot be used to fabricate entries under another person’s identity.
3. Virtual Assistants & Chatbots
Voice assistants act on what they hear — FARx ensures they only act for the right person. By verifying the speaker’s identity before commands are executed, FARx prevents unauthorised access to accounts, data, or sensitive functions. Deepfake detection stops attackers from using cloned audio to trigger commands or impersonate a registered user.
4. Accessibility
Voice-controlled accessibility tools must be both open and secure. FARx ensures only the registered user can operate the system — without adding friction or compromising the accessibility experience. Synthetic voice screening prevents cloned voices from being used to impersonate a registered user and gain unauthorised control.
5. Financial TradingOn trading floors, voice instructions carry significant financial and regulatory weight. FARx continuously verifies the identity of the trader throughout the session, creating an auditable record of who placed each instruction. Deepfake and synthetic voice detection guards against AI-generated clones being used to execute fraudulent or unauthorised trades.
6. Real-Time Transcription & Subtitling
Transcription captures the words; FARx confirms who spoke them. In meetings, court proceedings, or broadcasts, FARx enables speaker-attributed transcripts where each contribution is tied to a verified identity — not just a name that someone provided. Synthetic voice detection flags any AI-generated or cloned audio in the session, protecting the integrity of the record.
7. Automotive
In-car speech recognition responds to whoever speaks. FARx ensures it only responds to the registered driver — relevant for fleet operators, insurers, and connected vehicle platforms where accountability matters. Synthetic voice detection prevents cloned or AI-generated audio from being used to manipulate vehicle systems remotely.
8. Healthcare
Healthcare speech recognition captures clinical notes, EHR entries, and consultation recordings. FARx verifies that it is the named clinician speaking — ensuring records are correctly attributed and patient data access is controlled. Deepfake and synthetic voice detection prevents AI-generated impersonation from being used to access clinical systems or manipulate records.
9. Education
Speech recognition evaluates and records what students say. FARx confirms it is actually the enrolled student saying it — critical for remote assessments and spoken language evaluations where impersonation is a growing risk. Synthetic voice detection identifies submissions generated using AI voice tools, protecting the integrity of results.
10. Security & Surveillance
Voice monitoring detects what is being said in regulated or high-security environments. FARx adds the critical dimension of who is saying it — matching speakers against known individuals or watchlists in real time. Synthetic voice detection ensures that AI-generated or cloned audio does not evade screening, closing a significant gap in voice-based surveillance and intelligence operations.

Discover how FARx detects deepfakes, impersonation and identity fraud in real time.
Deepfake Detection