
Google Video creation post-I/O — early impressions and limitations
Just posted a walkthrough of what I found interesting from I/O for video creators. Quick notes:
- Flow Agent mode step-by-step agentic video workflow with feedback loop; promising structure, but safety filter triggers and script adherence are rough.
- Omni video model video-to-video editing is real and usable (restyling, object replacement, etc.) So cool!
- Flow Music (formerly producer.ai) now generates lipsynced performer video alongside audio
Still early on the agent side — curious if others have found workarounds for safety triggers or better ways to keep agents on script