AI and Interpretation Disclaimer

The service does not replace a qualified human interpreter. It may miss words, confuse speakers, choose the wrong language, mistranslate meaning, flatten tone, mishandle cultural context, invent connecting language, or speak an output that differs from the source.

Interim subtitles are provisional. The system may revise earlier words after receiving more context. Confirmed text and translated output may still contain errors. Do not treat an early subtitle as final merely because it appeared on screen.

Network conditions, microphone quality, background noise, browser behavior, provider availability, language pair, rate limits, device performance, and fallback paths can cause delay, dropped speech, repeated output, interruption, or silence.

“Real-time,” “continuous,” and “low latency” describe the intended experience, not a guarantee that every sentence will be delivered without delay.

Generated speech may not preserve the speaker's identity, emotion, emphasis, politeness, gender expression, accent, pacing, or intent. The audible voice is synthetic output and should not be represented as the original speaker's actual voice unless that is clearly true.

Names, dates, addresses, prices, quantities, medication names, legal terms, instructions, consent, and commitments should be repeated and independently verified. Participants should ask for clarification whenever the result is unclear or consequential.

Do not use Live Interpreter as the sole basis for emergency, medical, legal, immigration, financial, law-enforcement, educational-accommodation, workplace-safety, or other rights- or safety-critical communication. Use a qualified human interpreter and appropriate professional support.

A language shown in a menu does not guarantee equal accuracy, bidirectional support, automatic detection, dialect coverage, voice availability, or compatibility with every processing engine. Some modes may be single-direction or unavailable with certain provider combinations.

If predictive or context-assisted features are enabled, they may guess likely upcoming language before the speaker finishes. A prediction is not a transcript and may be wrong. The product should visibly distinguish predicted, interim, confirmed, and translated content.

You are responsible for deciding whether the service is appropriate for a conversation, obtaining required consent, monitoring the displayed result, stopping use when errors create risk, and confirming important meaning directly with the other participant.

Report repeated recognition, translation, latency, speaker-direction, or voice-output problems to [email protected]. Avoid sending unnecessary sensitive conversation content.