Frequently Asked
Questions

Aurigin.ai is a deep-tech security company that protects organizations from AI-generated voice fraud and deepfakes.

Our technology detects whether an audio sample, from a phone call, video meeting, or recording, is human or AI-generated, in real time and across 40+ languages.

We help banks, telcos, cybersecurity providers, and law enforcement agencies secure their communications and verify audio authenticity before damage occurs.

Aurigin.ai can be used through our desktop app, API, or on-premise deployments, giving enterprises full flexibility and data control.

Aurigin.ai offers flexible deployment options to fit different security and infrastructure needs:

Cloud (Aurigin Servers): The fastest way to get started. Process audio through our secure cloud API or desktop app, with no setup required.

On-Device (Desktop Apps): Run detection locally on Windows or macOS for full control over data and real-time monitoring during calls.

On-Premise: For enterprises with strict data policies, we provide on-premise deployment options within your own infrastructure or private cloud environment.

We take data privacy very seriously and give users full control over how their audio is handled.

Free Plans: Audio data may be collected to help us improve model performance and service reliability.

Paid Subscriptions: No audio data is collected or stored—your files are processed securely and immediately deleted after analysis.

On-Device or On-Premise Deployments: No audio data ever leaves your device or your organization's private infrastructure.

Across all plans, we collect usage and performance metrics (e.g. anonymized logs, feature use, errors) to maintain and improve the service—but never your audio content itself unless explicitly stated.

All data is processed in Europe and complies with GDPR and Swiss data protection standards.

3 seconds of audio are required to generate a reliable deepfake detection result.

Shorter clips typically don't contain enough speech patterns or acoustic features for the model to analyze authenticity with confidence. For best results, we recommend using audio segments of at least 5-10 seconds whenever possible.

Aurigin.ai works across 40+ languages and is language-agnostic, meaning it analyzes the audio signal itself, not the spoken words.

The model performs consistently across different accents, dialects, and recording conditions.

The percentage displayed on the Aurigin.ai web app represents how much of the analyzed audio file has been flagged as AI-generated. It is not a confidence score—rather, it shows the proportion of the audio our system identified as synthetic versus human.

If you use the API, you'll receive a more detailed output that includes a confidence score for each segment and for the file overall.

Aurigin.ai's detection engine operates with a model latency of under 50 milliseconds, enabling near real-time analysis.

Depending on your network speed and the file size being processed, results are typically available within a few seconds per minute of audio—even for longer recordings.

The upload size limit depends on how you use Aurigin.ai:

• Web App: Up to 30 MB per file

• API: Up to 4 MB per request (larger files can be split into smaller chunks)

• Desktop Apps: No size limit—audio is streamed and analyzed in real time as conversations occur

These limits ensure fast processing and optimal performance across all use cases.

Not yet. Aurigin.ai is currently available as a desktop application (Windows and macOS) and as an API for integration into your own systems.

We don't offer a mobile app for iOS or Android at this time. However, mobile compatibility is part of our long-term roadmap as we expand into new use cases and platforms.

At the moment, Aurigin.ai focuses exclusively on audio deepfake detection, identifying whether a voice or sound recording has been synthetically generated or manipulated.

Our technology is optimized for real-time voice analysis with high accuracy and low latency across 40+ languages.

Not yet. Aurigin.ai currently focuses on detecting whether an audio sample is AI-generated or human-made.

We don't yet detect replay attacks. However, replay and injection attack detection is part of our upcoming roadmap, as we continue expanding from deepfake detection to broader voice fraud protection.

Yes. Aurigin.ai allows enterprises to fine-tune detection thresholds to balance false positives (real audio flagged as AI) and false negatives (AI audio classified as real) based on their risk tolerance and use case.

This calibration can be applied through the API, on-premise, or custom enterprise deployments, enabling optimal performance for your specific operational context.

Aurigin.ai is currently undergoing SOC 2 Type II and ISO 27001 certification processes to validate our commitment to the highest standards of information security and data protection.

While the audits are in progress, all our deployments already follow the same security controls and procedures required by these frameworks.