Tim Simmons tests Seed Audio on dialogue and sound effects, then compares video generations made with and without an audio reference. He highlights differences between examples and notes that a preferred visual result is not guaranteed.
Simmons experiments with dialogue scenes, using image references, timestamps and light audio editing to address repeated lines and other inconsistencies. He notes that audio generations often run longer than the desired shot and may need trimming.
He demonstrates a separate camera-motion workflow for moving through a generated scene without manually building simplified geometry in Blender. His observations describe the tested examples rather than a controlled comparison of the tools.
Watch on YouTube
