video1.0<iframe src="https://www.loom.com/embed/a6af3e55e93246b79cf71a28d54e96cd" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/a6af3e55e93246b79cf71a28d54e96cd-1ac420e390ecc9d0.gif167.915This Loom demonstrates Cinebose’s real-time voice AI running on a phone with an offline, fully on-device mode and a cloud fallback mode. The speaker explains that in fully on-device mode, speech-to-text, text-to-speech, and a small language model run directly on the phone, while in cloud fallback mode only text is sent to the cloud, with speech-to-text and text-to-speech still on-device. They claim the cloud fallback can be about 99% cheaper and potentially much faster with optimization. The demo shown includes an offline travel-style assistant that responds with taxi cost estimates and suggested local places to visit in Barbados.