Overview
Google DeepMind has released Gemini Omni 1.2 Flash, a video generation and editing model available through the Gemini API, Google Cloud, Google Vids, and Gemini for business. Google claims the model produces native clips up to 15 seconds long, accepts character, product, or style images alongside motion clips and audio cues, and allows clips to be refined across multiple conversational turns without regenerating them from scratch [1][2].
Pricing for 720p output is $0.07 per second, a 30% reduction from Omni 1.1. Google positions the model for product explainers, social ad spots, virtual try-on sequences, catalog spins, and other short-form commercial video [1][2].
The update is primarily a vendor-defined capability and price change. Independent benchmarks, user case studies, and comparative analysis with competing video models are not available, so real-world quality, latency, and adoption remain unknown.
Context
Gemini Omni 1.1 Flash, the preceding version named in Google's announcement, supported video clips up to 10 seconds. Omni 1.2 Flash raises that ceiling to 15 seconds and adds selectable durations in one-second increments from 3 to 15 seconds, allowing a single generation to cover a complete short scene rather than requiring multiple clips to be assembled [1][2].
Google's proposed workflow separates cheap drafting from final output: creators can draft at 360p or 720p, apply a targeted conversational edit, and then upscale selected clips to 1080p or 4K. The company says editing a detail through a follow-up prompt avoids regenerating the entire clip, but it does not publish the prices for 1080p or 4K output, so the total cost of that workflow is not documented [1][2].
Why it matters
The combination of a longer single-generation length and turn-by-turn editing suggests a shorter production loop for short commercial video. Product explainers, social ad variants, and catalog spins that previously required several generations could be handled as one 15-second clip with subsequent refinements [1][2]. This is a reasoned inference rather than a measured outcome; Google does not report regeneration rates, time saved, or comparison data.
Reference handling is the second substantive change. Google claims that Omni 1.2 Flash keeps faces, garments, packaging typography, and brand colors consistent across camera angles, and that it follows reference images more closely than Omni 1.1 [1][2]. If that holds in routine use, it would make the model more relevant to brand and product teams whose outputs depend on identifiable details staying stable across a shot. The vendor's example includes a 15-second virtual try-on sequence across 360-degree turns using a model photo and front-and-back images of three outfits, with the model's identity and garment details reportedly preserved [1][2].
The 30% price cut to $0.07 per second for 720p lowers the cost of repeated drafts for smaller teams and independent creators [1][2]. Whether that translates into cheaper final production depends on factors the announcement leaves unaddressed, including higher-resolution pricing, latency, and how often edits trigger regeneration.
Opportunities
- Short-form commerce video: Product teams could generate 15-second product explainers, catalog spins, and virtual try-on sequences from existing photography, using conversational edits for placement-specific ad variants [1][2].
- Drafting at lower resolution: A workflow of 360p or 720p drafts, conversational correction, then 1080p or 4K upscaling could reduce the amount of final-resolution generation, though the high-resolution price is not disclosed [1][2].
- Multi-variant production: Multi-turn editing may let publishers produce several background, lighting, or aspect-ratio variants from one initial clip without re-briefing [1][2].
Risks
- Unverified consistency: All quality comparisons come from Google's own examples, including the claim that Omni 1.2 Flash matches references more closely than Omni 1.1; no independent test results or user reports confirm the improvement in routine use [1][2].
- Incomplete pricing: The $0.07 per second figure applies to 720p only. Costs for 1080p and 4K, and any per-edit charges, are not stated, so final budgets cannot be calculated [1][2].
- Platform dependency: The model is available through Google's Gemini API and Google Cloud, which concentrates production workflows in a proprietary service where pricing and availability can change [1][2].
- Unmeasured operational behavior: Latency, scalability, hardware requirements, and the rate at which edits trigger new generation are not documented [1][2].
- Reference rights: Using copyrighted images, audio, or voice references in generated commercial video introduces rights questions that the announcement does not address.
Evidence & sources
The key statements of this analysis, labeled by type, with the sources they rely on. Trendoviz does not run product tests.
- Vendor statementGoogle claims Gemini Omni 1.2 Flash generates and edits video natively up to 15 seconds, compared with 10 seconds in Omni 1.1.[1][2]
- Vendor statementGoogle says clip duration can be set in one-second increments from 3 to 15 seconds.[1][2]
- Vendor statementGoogle reports a 30% price reduction from Omni 1.1, bringing 720p video to $0.07 per second.[1][2]
- Vendor statementGoogle claims multi-turn conversational editing lets users revise the same clip across its full 15-second length without regenerating it from scratch.[1][2]
- Vendor statementGoogle says generation can be guided by character, product, or style images, motion clips up to three seconds, and audio cues.[1][2]
- Vendor statementGoogle claims the model holds faces, garments, packaging typography, and brand colors consistent across camera angles, with less visual drift than its predecessor.[1][2]
- HypothesisA 15-second single-clip workflow with conversational revisions could shorten the production loop for product explainers and ad variants.[1][2]
- Vendor statementThe model is generally available through the Gemini API, Google Cloud, Google Vids, and Gemini for business.[1][2]
- [1]Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips●deepmind.googleretrieved Nov 18, 2026
- [2]Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips●blog.googleretrieved Nov 18, 2026
Signals
- Nov 5Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips●deepmind.google
- Nov 5Gemini Omni 1.2 Flash: Fast and cost-effective with longer clips●blog.google
Timeline
Analysis
The evidence for this trend consists of two official Google announcements describing the same release, so it supports a clear account of what the vendor says the model does, but not of how it performs outside Google's own examples [1][2]. The concrete, checkable changes are the extension of native clip length from 10 to 15 seconds, selectable one-second durations between 3 and 15 seconds, multi-turn editing, general availability across Google platforms, and a 30% price reduction to $0.07 per second for 720p [1][2].
The consistency and motion claims are more qualitative. Google reports fewer unwanted jump cuts in continuous FPV movements, clearer lip-sync, and distinct multi-speaker conditioning in specific demonstrations with architectural photographs, character sheets, and one-second voice references [1][2]. These are illustrative scenarios selected by the vendor. A reasonable hypothesis is that improved reference adherence matters more than clip length for commercial use, because product and brand work depends on stable typography, colors, garment cut, and identity; the opposite is also possible, and no evidence here settles which effect dominates.
The pricing change is directly stated but incomplete in scope. The per-second rate makes short drafts inexpensive on their face, and the draft-then-edit-then-upscale workflow could avoid repeated final-resolution renders. That would reduce spend only if the edit path does not bill as a new generation and if higher-resolution output is priced reasonably, neither of which is documented. Overall, the release is well defined as a feature and price announcement and poorly defined as a measured quality improvement.
Limitations & uncertainties
The available material is limited to Google's own announcements, which provide vendor claims rather than independent evaluation [1][2]. There are no third-party benchmarks, side-by-side tests, or published performance measurements for clip quality, reference adherence, or motion continuity.
No user testimonials or case studies are included, so real-world reliability and satisfaction are unknown. Pricing is disclosed only for 720p; 1080p and 4K costs, and any fees for conversational edits, are not stated. Latency, throughput, hardware requirements, and scalability are not documented. There is no comparative analysis with competing video generation models, and the announcements do not describe SDK integration guides, developer support, availability by region, or usage limits.
Updates & corrections
- First published · published automatically after passing every check
Found an error? Report it.