Right now Antigravity only pulls from the text and code models like Gemini 4 Argon and 3.8 Flash. That works fine for standard dev tasks, but it breaks the workflow whenever a project needs media generation or audio checks.
We need native hooks for Google Flow and Gemini Omni right inside the workspace. If an agent is building out a front-end or a media pipeline, it should be able to call Omni or Flow directly to generate video assets, edit existing clips in-chat, or build scenes without me having to leave the IDE, open a browser, and manually move MP4s around.
Same problem with audio. Agents need a built-in function to analyze raw audio files directly. Right now if an agent runs a test or pulls an audio output from Omni, it can’t actually listen to the track to verify the sound design, check if the dialogue clipped, or confirm beat sync. Let the subagents ingest .wav and .mp3 files natively to check timestamps, frequency issues, and speech accuracy on their own.