Gemini Omni 1.2 Flash for Product Managers: 15-Second Video Generation Is Now Generally Available
TL;DR
Google released Gemini Omni 1.2 Flash on November 5, 2026. It generates 15-second videos from text, image, and video inputs, supports multi-turn conversational editing (ask it to change specific elements without regenerating everything), and ships SynthID watermarks on every output. Pricing is approximately $1.50 per million input tokens and $9.00 per million output tokens (verify against Google's current pricing page). The model is generally available via the Gemini API and Google Cloud. Teams using Gemini Omni 1.1 Flash should know their production traffic will auto-route to 1.2 Flash on December 3, 2026 per AI Studio release notes.
The AI PM Minute
One tactic to make you a sharper AI PM, twice a week. 60 seconds to read. Free.
No fluff. Unsubscribe anytime.
What Gemini Omni 1.2 Flash Is
Gemini Omni 1.2 Flash is a generative video model from Google DeepMind, the successor to Gemini Omni 1.1 Flash. It sits in the Gemini 3 series of models and handles native video creation and editing. Inputs can be text, images, or existing video clips. Output is video, with text descriptions also supported.
The name follows the Gemini Flash naming convention: Flash models are cost-optimized and fast, trading some output quality for speed and price compared to the full Omni model. Gemini Omni 1.1 Flash was already broadly used for video generation tasks in production. Version 1.2 extends the generation length to 15 seconds, adds multi-turn conversational editing, and improves reference following (how precisely the model matches a reference image or described visual style).
Video generation
Core capabilityCreates 15-second videos from text descriptions. Can incorporate reference images to anchor visual style, characters, or settings. Each generation includes an imperceptible SynthID watermark from Google.
Multi-turn conversational editing
New in 1.2After generating a video, you can send follow-up instructions: 'make the background warmer,' 'slow down the second half,' 'replace the character with the reference image.' Avoids full regeneration for iterative changes.
Enhanced reference following
Improved in 1.2Improved consistency between a reference image and the generated output. Useful for brand-controlled video generation where logos, colors, and characters need to match source assets.
Provisioned Throughput
EnterpriseAvailable for enterprise workloads that need guaranteed throughput rather than on-demand. Relevant if your product requires consistent video generation capacity rather than variable burst.
Product Use Cases: Where Video Generation Earns Its Place
Generative video has had a checkered product history. Early models were expensive, slow, and produced results that looked convincing in a demo but failed under real usage conditions. Gemini Omni 1.2 Flash is a different conversation: generally available API access, priced for integration-level use rather than research, and production-ready quality for the right use cases.
The right use cases share a few properties: the content length is 15 seconds or less, the output does not need to be photorealistic, and the value of fast generation outweighs the value of maximum quality. Here are the use case categories that fit well.
Social media and marketing content generation
15 seconds is exactly the format for short-form social video. Brands that produce high volumes of product videos, event teasers, or campaign variants benefit from programmatic generation at this length. The enhanced reference following makes brand consistency tractable.
Product walkthroughs and feature explainers
Short animated explainers of product features, onboarding steps, or how-it-works sequences. Faster and cheaper than commissioning animation for each feature update. Best for products where feature velocity is high.
Personalized video messages at scale
Generating personalized short video clips for different user segments or individual users. The multi-turn editing capability helps tweak personalization elements without full regeneration, but quality limits apply at the per-user level.
Long-form documentary or storytelling video
15 seconds is a hard ceiling on generation length. Long-form video requires stitching generations together, which introduces consistency problems. Use specialized long-form video tools for this use case.
High-fidelity visual effects or broadcast quality
Omni 1.2 Flash is a Flash model: cost and speed are optimized over maximum quality. For broadcast-grade or high-fidelity VFX needs, evaluate the full Omni model when available rather than the Flash tier.
Pricing and What It Means for Your Unit Economics
Third-party directories list Gemini Omni 1.2 Flash at $1.50 per million input tokens and $9.00 per million output tokens. Verify these against Google's current pricing page before building a business case: introductory pricing on new Google models has changed without announcement before. The pricing structure also reflects the multimodal nature of the model: a video input and a video output are both token-priced, not second-priced, so your cost depends on how the model tokenizes video at the resolution and length you use.
Input pricing
~$1.50 per million tokens. For video inputs, token count scales with resolution and duration. A 15-second HD video clip as input is substantially more tokens than a text prompt. Test your specific input format to get real cost figures.
Output pricing
~$9.00 per million tokens. Output video is priced on the generated token count. A 15-second output at standard resolution will cost more than a 5-second output at the same resolution. Benchmark your p50 and p95 cost per generation against your planned volume.
Provisioned Throughput
Available for enterprise workloads. Priced separately and provides guaranteed throughput. Relevant if your product generates video at scale and cannot tolerate on-demand latency or capacity variability.
SynthID watermark
Every output carries an imperceptible SynthID watermark. This is not removable. For products where AI origin disclosure is a legal or brand requirement, the watermark is a built-in compliance feature. For products where watermarks are unacceptable, evaluate whether this is a blocker.
Build Products With the Latest AI Capabilities
The AI PM Masterclass covers multimodal AI products, model selection, and agentic architecture in live instructor-led sessions with a Salesforce Sr. Director PM.
Migration from 1.1 Flash: What You Need to Do Before December 3
Per AI Studio release notes, Gemini Omni 1.1 Flash is scheduled to automatically route to gemini-omni-1.2-flash on December 3, 2026. This is a standard Google migration path: the previous Flash model stays listed as "stable" but begins serving the newer version under the hood on the migration date.
Before December 3, run your eval suite on Omni 1.2 Flash explicitly. Even minor version changes can shift output behavior enough to affect quality gates, latency expectations, or specific generation characteristics your product depends on. The goal is not to prevent the migration but to ensure it does not surprise you in production.
1. Identify all 1.1 Flash calls in your codebase
Search for 'gemini-omni-1.1-flash', 'gemini-omni-flash', or any alias that resolves to Omni 1.1 Flash. These are the calls that will auto-migrate. List the production use cases each serves.
2. Run your eval suite on 1.2 Flash before December 3
Point your test harness at gemini-omni-1.2-flash explicitly and run your standard quality checks. Focus on: generation quality at your target resolution, edge-case prompts, and latency at your p95 load.
3. Test multi-turn editing if you plan to use it
Multi-turn conversational editing is new in 1.2 Flash. If your product flow could benefit from iterative video refinement, evaluate it now. It changes how you structure generation requests.
4. Update your model ID pinning if needed
If your eval finds regressions, pin explicitly to gemini-omni-1.1-flash until you resolve them. The auto-migration happens on alias-based calls; pinned version calls will continue on 1.1 Flash until 1.1 is retired.
Decision Framework: Should You Build on Gemini Omni 1.2 Flash?
Video generation APIs have crossed a threshold in 2026: they are fast enough, cheap enough, and reliable enough for integration-level use in products that benefit from them. The question is whether your product is one of them.
Generative video belongs in a product when the generation result is good enough to replace or augment a human-produced video at a cost and speed that changes the product's economics. It does not belong when users will immediately see the AI origin as a quality problem, or when the content requirements exceed what 15-second generation can produce.
Build on it now
- •Your product generates short-form marketing, social, or explainer video at scale
- •You are already using Gemini Omni 1.1 Flash and want the multi-turn editing capability
- •Brand consistency in generated video matters to your users (enhanced reference following helps)
- •Your users need fast video output and 15 seconds covers the majority of use cases
Evaluate before committing
- •Your use case is at the edge of the 15-second limit and you need to test whether clip-stitching degrades quality unacceptably
- •Your product serves a brand or creative context where SynthID watermarking needs legal review
- •You are evaluating multiple video generation APIs and need head-to-head quality comparison
Not the right model for this use case
- •Your content requires longer than 15 seconds with narrative continuity
- •You need broadcast-grade or VFX-quality video output
- •Your product is a video platform where AI origin detection would undermine user trust
Ship AI Products With Confidence
The AI PM Masterclass covers multimodal AI products, model evaluation, and production architecture in live instructor-led sessions. Taught by a Salesforce Sr. Director PM.
Related Articles
Before you go: get the AI PM Minute
One tactic to make you a sharper AI PM, twice a week. 60 seconds to read. Free.
No fluff. Unsubscribe anytime.