Gemini Omni 1.2 Flash
Google’s release of Gemini Omni 1.2 Flash, a paid video-generation model offering up to 15‑second clips and lower API pricing, is met with mixed reactions: some see real potential for creative and B‑roll-style workflows, while others question the value and societal impact of ever-more-convincing synthetic video. Many commenters focus less on raw model quality and more on Google’s confusing branding and product matrix, fragmented rollout, and notoriously bad developer and subscription UX that make the models hard to even access. Concerns also surface around lock-in, reliance on personal Google accounts, and perceived neglect of more practical integrations, like better AI support in Workspace apps.
Model naming & positioning
- Many find “Gemini Omni 1.2 Flash” confusing against “Gemini 3.x/4” and existing names like Veo, Flash, Pro, Ultra, Argon, Carbon.
- Some want car‑style or year‑based versions (e.g., “Omni 2026.2”) or simple capability tiers (low/medium/high).
- Confusion over “Omni” for a video-only model when “omni” sounds like general multimodal; suggestion to just call it “Gemini Video”.
Pricing & capabilities
- Omni 1.2 Flash offers 15‑second native video generation, video extension, multi‑turn editing, and supposedly better character consistency.
- API pricing mentioned as $0.07/second for 720p and cheaper than Omni 1.1 via API, but in Google Flow it appears more expensive per clip.
- Some see 15 seconds as mainly useful for B‑roll; others emphasize movies are built from many short shots, so longer coherent clips matter.
Access, rollout & UX
- Recurrent complaint: extremely confusing access paths across Gemini app, Google One AI plans, Workspace, Flow, Vids, AI Studio, Vertex, and Antigravity/CLI.
- Rollout appears staggered and opaque: different limits and timelines for free, Pro/Ultra, business, and API users.
- AI Studio and broader Google Cloud UX are heavily criticized as convoluted and brittle; some users can’t even get billing or model selection to work.
- A few say access “just works” via certain subscriptions or third‑party routers (OpenRouter, Vercel), leading others to suspect big variability by account type/region.
Comparisons to other providers
- Several prefer Anthropic/OpenAI (and open‑weight models) because they are easier to sign up for, have clearer plans, and more consistent harnesses.
- Others report Google models (especially newer ones like 3.8 Flash/Argon) are competitive or better for some coding or writing tasks, but hampered by limits and tooling.
Use cases & adoption of video models
- Suggested uses: ad/b‑roll creation, low‑budget filmmaking, driver fatigue datasets, and special‑effects shots that would be unaffordable otherwise.
- Skepticism about long‑term viability: poor adherence to detailed prompts, trouble with atypical character designs, need for many iterations.
Trust, safety & product direction
- Some fear tying AI use to personal Gmail due to the risk of account bans.
- Complaints that Google prioritizes flashy demos and search/ads integration over robust, developer‑friendly products (e.g., weak Gemini in Docs/Slides vs strong standalone competitors).
- A minority note that mainstream users do engage heavily with Gemini‑powered search summaries, even if that unit is reportedly unprofitable.