Gemini Omni 1.2 Flash is GA: 15-second clips, multi-turn edits, ~$0.07/s at 720p — and 1.1 routes to it on Dec 3

Google DeepMind has shipped Gemini Omni 1.2 Flash (gemini-omni-1.2-flash) to general availability on the Gemini API and Google Cloud. Product manager Anish Nangia's November 5 post and Logan Kilpatrick's rollout note frame it as a production video model: longer clips, cheaper output, and edits that stay in the same conversation instead of regenerating from scratch.

What actually changed from Omni 1.1

Per the model page and the blog:

• Native length: generate and edit up to 15 seconds (was 10s on 1.1), in one-second steps from 3–15s. First/last-frame interpolation and per-turn extension stay in the 3–10s band; extensions can stack to about 30 seconds total.
• Price: Google says video output is 30% cheaper than Omni 1.1 — $0.07 per second at 720p on the paid API (blog figure). Draft at 360p/720p, then upscale keepers to 1080p or 4K.
• References: up to six reference images per request (blog/docs), plus short motion clips and audio cues, with stronger claims about face, garment, packaging and brand-color consistency across angles.
• Multi-turn editing: refine the same clip in conversation — swap a product, change lighting or camera — without starting a new generation.
• Speech: Google claims more natural dialogue, clearer lip-sync and better multi-speaker conditioning. Audio references remain allowlist-gated on Cloud.

Outputs carry SynthID. The DeepMind model card (published 5 November 2026) says Omni 1.2 Flash does not add meaningful new frontier-safety capabilities beyond Gemini 3.1 Pro / 3.7 Flash assessments.

Rate limits and the Dec 3 cutover

Paid-tier interactive limits for gemini-omni-1.2-flash, from the rate-limits docs: Tier 1 8 RPM / 40 RPD, Tier 2 16 / 200, Tier 3 40 / 2,000. Free tier is unsupported. Google's own release notes say paid-tier limits are doubled versus the previous Omni Flash table; treat that as Google's claim against the numbers above.

Developers still on gemini-omni-1.1-flash get a short runway. Effective December 3, 2026, those requests auto-route to 1.2 and bill at 1.2 rates — no code change required, but the cheaper (and different) model is what answers. The pattern matches Google's other ID redirects; we covered a related December cutover for Gemini 2.5 → 3.x Flash IDs in an earlier post.

Editorial note

This is a product increment with numbers you can check: length, price per second, reference count, rate-limit table, and a hard routing date. The consistency and "natural speech" claims are Google's; independent side-by-sides against Omni 1.1 and against rival video APIs are still the missing piece. Google itself flags remaining weakness on fast sports action, rapid script changes and complex physics.