AI Voice & Media Disclosure

1. Autonomous Presenter Declaration

Aura is an autonomous artificial intelligence agent, not a human presenter. Aura dynamically reads slide presentations, synchronizes visual decks via finite-state machine orchestration, and converses with attendees through synthesized voice audio in real time. In compliance with global transparency standards (including Article 52 of the EU AI Act and FTC guidelines), Seminara provides clear indicators within presentation rooms confirming that attendees are communicating with an AI system.

2. Real-Time Voice Pipeline Architecture

Aura’s conversational audio loop operates over high-performance WebRTC peer connections designed to achieve human-paced interactive latencies (<800ms round-trip):

• Audio Ingestion: Attendee microphone audio is packetized into Opus frames and relayed via encrypted WebRTC edge gateways.
• Acoustic Speech Recognition: Audio is converted to text tokens in real time under zero-retention commercial API parameters.
• Contextual Reasoning: Aura determines whether to continue presenting the slide narrative or pause to answer an attendee’s verbal question using the host’s private knowledge base.
• Neural Speech Synthesis: High-fidelity synthetic speech audio is streamed back through WebRTC to all connected room participants.

3. Ephemeral Audio vs. Transcript Retention

Raw voice audio streams are processed ephemerally in server volatile memory (RAM). Seminara does not record, archive, or persist raw audio files (such as MP3 or WAV files) of attendee speech to disk storage. Text transcripts of attendee questions and Aura’s answers are archived in the host’s private workspace under the host’s sole control to export or permanently purge.

4. Non-Biometric Processing Guarantee

Seminara does not create, extract, capture, store, or analyze biometric identifiers or voiceprints from attendee speech (in compliance with the Illinois BIPA, Texas CUBI, and Washington Biometric Privacy Law). Our speech recognition models process acoustic audio signals strictly to generate linguistic text strings.

5. Host Guidelines for Attendee Audio Notice

Under electronic surveillance and two-party wiretapping laws (such as California Penal Code § 632 and GDPR Article 6), participants must be informed when their voice communication is intercepted and transcribed. Hosts must ensure that any invitation link, calendar invite, or registration page states clearly:

“This presentation features Aura, an autonomous AI presenter. By unmuting your microphone to participate in live voice Q&A, you acknowledge that your audio is processed in real-time to facilitate interactive discussion.”

6. Voice Ethics & Anti-Deepfake Standards

All synthetic voices utilized by Aura are commercially licensed professional voice models crafted specifically for conversational software. We enforce strict safeguards preventing the unauthorized cloning of voices of living or deceased individuals without verified contractual authorization.

7. Latency, Speech Disclaimers & Bounds

While Aura’s conversational engine is engineered for low latency and high factual adherence to the host’s uploaded slide deck, participants should note that network packet loss, public Wi-Fi jitter, and foundational generative model behaviors can occasionally cause conversational pauses, brief turn-taking interruptions, or unintended spoken statements. Hosts remain responsible for reviewing slide facts and answering edge-case customer queries.

8. Technical Inquiries & Contact

For technical questions regarding our WebRTC audio architecture or enterprise synthetic media compliance, reach our engineering desk at shivamselam@seminara.online.