Table 1. I2V samples with scene cuts generated by FrameBridge.
Sample 1: "Car driving on the road. Smoothly changed to the sunset" Sample 2: "Camera zooms in. Smoothly changed to another dish in the restaurant"
Condition Image: I2V Result: Condition Image: I2V Result:
Table 2. I2V samples with scene cuts generated by FrameBridge.
Sample 3: "Cars driving on the road. Vertically fade in to a house in the forest" Sample 4: "The waves crash against the shore. Vertically changed to another scene of ship sailing in the sea"
Condition Image: I2V Result: Condition Image: I2V Result:
Table 3. I2V samples with scene cuts generated by FrameBridge.
Sample 5: "Camera zooms out. Abruptly changed to a garden with birds flying" Sample 6: "Leaves blown off by the wind. Abruptly changed to a streaming river at a different time and place"
Condition Image: I2V Result: Condition Image: I2V Result:
Table 4. Example: Sampling process of bridge models from $z_T$ to $z_0$. We select 8 intermediate timesteps $0=t_1 < t_2 < .. < t_7 < t_8 = T$
$z_{t_8}$: $z_{t_7}$: $z_{t_6}$: $z_{t_5}$:
$z_{t_4}$: $z_{t_3}$: $z_{t_2}$: $z_{t_1}$: