Each image in a podcast episode video can now be linked to one of the people who recorded the episode. While that person is speaking the image is shown at its speaking opacity; while they are quiet it drops to the quiet opacity, so a picture of the current speaker stands out without them all flickering on and off.
Open the video editor for a recorded episode, select an image, and choose the participant under React to speaker. Set the two opacities if the defaults do not suit. Leave it on None for an ordinary static image.
Speech activity is worked out from the participant's own isolated recording when the take is prepared, so creating the video does not re-analyze the audio. Several people talking at once means several images are bright at the same time. Episodes recorded before this change still export as before.