third-eye / DEMO_SCRIPT.md
mitvho09's picture
Upload folder using huggingface_hub
031e3f9 verified
|
Raw
History Blame Contribute Delete
1.11 kB

A newer version of the Gradio SDK is available: 6.19.0

Upgrade

Third Eye: 50-second demo

Time Picture Audio
0-5s Black screen. Iris opens and begins its idle glow. "What if your phone could see for you?"
5-14s First-person view. Select the bundled restaurant menu. Room tone only.
14-25s Select Chinese and tap Read this text. Iris moves through Seeing and Thinking. Third Eye summarizes the menu in Chinese.
25-34s Cut to the medicine label. Large transcript appears. "Amoxicillin, 500 milligrams. Take one capsule every eight hours."
34-43s Cut to the station sign. Use Ask mode: "Which way is the station?" "Central Station is 250 meters to the left."
43-50s Iris speaking state, then title card. "Third Eye. Built on a 2.8 billion parameter visual model. Designed to move on-device."

Recording notes

  • Record a real end-to-end run; do not splice a fake model answer over the interface.
  • Keep the answer transcript visible whenever audio plays.
  • Show the language selector before the Chinese segment.
  • End card: OpenBMB, Cohere Labs, Modal, Gradio, Hugging Face.