Google media model ids on Oct 4: Omni, Veo 3.1, Lyria 3.5 vs Sume ids

On October 4, 2026 the Google ids to use are gemini-omni-1.1-flash for video and lyria-3.5 for music. The three Veo 3.1 preview ids shut down on October 22, 2026, with Omni named as the replacement. On Sume, Omni is gemini-omni-flash-1.1 on the Video Router and Lyria 3.5 is lyria-3.5 on the Music Router. Sume's catalog has no Veo, so a Veo job migrates to Omni or another router model.

The crosswalk

Status comes from Google's models page and deprecations page, and the dates from its release notes. Sume ids come from the Video Router and Music Router docs.

Traps in the table

The Omni id word order differs between Google (omni-1.1-flash) and Sume (omni-flash-1.1). A copy-paste across the two fails on an unknown model. Google's Omni runs through the Interactions API while Veo used a different call shape, so migrating is a rewrite, not a rename.

On the music side, Sume's sume/music-auto lets the router pick; the response's job.request.routed_model names the engine used. Sume's Music Router rejects a duration field and a negative_prompt, and bills a fixed $0.125 per generation, so compare cost per finished track using Google's current pricing page rather than assuming the two meters match.

A migration order

Move Veo jobs first, because they have a hard date. Replace the id, rebuild the request for Omni's task values (text to video, image to video, reference to video, edit, extend), rerun your golden prompts, and compare cost per finished second. Then move music if you still call a Lyria preview. If you go through Sume instead, the single Video Router request shape means the switch is a model string, but you give up Veo itself.