Google Makes AI Video Generation 30% Cheaper With New Gemini Omni 1.2 Flash Model

Google has introduced Gemini Omni 1.2 Flash, a new artificial intelligence video-generation model that reduces video creation costs by 30% while offering longer clips, improved visual consistency and conversational editing capabilities.

Announced on November 5, 2026, the model is designed to make AI-powered video production more affordable for developers, businesses and content creators. It can generate videos up to 15 seconds long, compared with the previous 10-second limit, and allows users to refine existing clips through follow-up instructions.

One of the biggest changes is the reduction in 720p video-generation pricing to $0.07 per second through the Gemini API. This means a 15-second video costs approximately $1.05 in generation charges, before any additional processing or other applicable costs.

The launch could make AI-generated advertisements, product demonstrations, social media videos and promotional content more accessible to businesses that regularly produce short videos.

Gemini Omni 1.2 Flash: What’s New?

Gemini Omni 1.2 Flash is Google’s latest video-generation model in the Omni series. It combines video creation with editing capabilities that allow users to make changes through natural-language instructions.

Instead of generating an entirely new video whenever a small detail needs to be changed, users can provide additional instructions to modify an existing clip.

For example, a business could generate a product advertisement and then ask the model to change the background, lighting, camera angle or surrounding environment while attempting to preserve the main product and its appearance.

Google says the updated model also follows image and video references more accurately, helping maintain consistency in characters, products, clothing and brand details across different shots.

The main improvements include:

• 30% lower 720p video-generation pricing.
• Video generation lasting up to 15 seconds.
• Multi-turn conversational video editing.
• Better consistency for characters and products.
• Improved reference-image accuracy.
• Draft generation at 360p and 720p.
• Upscaling options to 1080p and 4K.
• More natural dialogue and improved lip-sync.

These improvements are intended to make the model more useful for both individual creators and organisations producing videos at scale.

Google Reduces AI Video Generation Price by 30%

The price reduction is among the most significant parts of the announcement.

Google has reduced the reported 720p video-generation cost from $0.10 per second with Gemini Omni 1.1 Flash to $0.07 per second with Omni 1.2 Flash.

The difference becomes more noticeable when generating multiple videos or longer clips.

Calculations use the reported per-second 720p API rates. They are illustrative and do not include additional costs or account-specific charges.

For example, generating a 15-second video at the new rate costs approximately $1.05, compared with $1.50 at the previous rate.

That represents a saving of $0.45 for each 15-second generation.

For businesses producing hundreds of promotional videos, product clips or advertisements, lower per-second pricing could reduce production expenses.

However, actual savings will depend on the number of generations, editing requirements, output resolution and overall workflow.

Videos Can Now Be Up to 15 Seconds Long

Another major upgrade is the increased video duration.

Gemini Omni 1.2 Flash supports video generation from three seconds up to 15 seconds, with duration adjustable in one-second increments.

The previous Omni 1.1 Flash model supported clips of up to 10 seconds.

The additional five seconds may appear modest, but it gives creators more room to communicate a product feature, demonstrate a service or build a short visual story.

For example, an online retailer could use a single generation to show a product from different angles, highlight its main features and finish with a promotional scene.

Similarly, a social media creator could produce a short video containing an opening visual, a central action and a concluding shot.

Longer clips may also reduce the need to combine several separately generated scenes, although the final result will still depend on the model’s ability to follow the prompt.

Conversational Editing: Change Videos Without Starting Again

One of the more useful additions is multi-turn conversational editing.

This allows users to continue refining a generated video by providing follow-up instructions.

For example, a user could begin with a prompt asking for a product advertisement featuring a smartphone on a desk.

After the first video is generated, the user could request:

• Replace the background with a modern office.
• Change the lighting to a warmer tone.
• Move the camera closer to the product.
• Adjust the camera angle.
• Change the weather outside the window.
• Modify the surrounding environment while preserving the main subject.

The model attempts to apply these changes while maintaining important visual elements from the original clip.

This approach could be particularly useful for advertising agencies, online stores and content production teams that regularly need different versions of the same video.

However, conversational editing does not guarantee that every detail will remain unchanged. Users should review the resulting footage for visual errors, inconsistent objects and unintended changes.

Improved Character and Product Consistency

Maintaining consistency between video scenes has been a persistent challenge for generative AI tools.

A character may look different from one shot to another, while product packaging, colours and other visual details can change unexpectedly.

Google says Gemini Omni 1.2 Flash improves reference adherence to help address these problems.

The model can use several types of references, including:

Character images: Users can provide photographs to guide the appearance of people or characters in generated videos.

Product images: Businesses can use product photographs to help maintain the appearance of items across different shots.

Style references: Creators can provide visual examples to guide the overall appearance and style of a video.

Motion references: Short video clips of up to three seconds can help guide movement.

Audio cues: Audio references can also be incorporated in supported workflows, although access to this functionality is restricted through allowlisting.

These capabilities could help businesses create product demonstrations, catalogue videos, virtual try-on experiences and branded advertisements using existing visual assets.

For example, an online clothing retailer could provide a model photograph and images of different outfits. The model could then generate a video showing the clothing from multiple angles while attempting to preserve the model’s appearance and garment details.

More Natural Dialogue and Improved Lip-Sync

Google has also highlighted improvements in generated speech and audio-related video features.

Gemini Omni 1.2 Flash is designed to produce more natural-sounding dialogue, clearer lip synchronization and distinct voices in scenes involving multiple speakers.

These improvements could be useful for:

• Short promotional advertisements.
• Product explainers.
• Dialogue-based social media videos.
• Character-based storytelling.
• Marketing campaigns.
• Educational content.

More accurate lip-sync can make generated characters appear more natural when speaking.

However, creators should still review generated speech for pronunciation, timing, factual accuracy and audio quality before publishing videos commercially.

Draft Videos at 360p or 720p, Then Upscale to 4K

Another feature aimed at reducing production costs is the ability to create lower-resolution drafts before producing a higher-resolution final version.

Users can generate drafts at 360p or 720p, review the results and make changes before upscaling selected videos.

The workflow can be useful because creators do not necessarily need to produce every early version at the highest resolution.

A possible production process could look like this:

1. Generate an initial video draft at 360p or 720p.
2. Review the composition, movement and visual details.
3. Make corrections through follow-up instructions.
4. Select the best version.
5. Upscale the selected output to 1080p or 4K.

This approach could help creative teams avoid spending higher-resolution production resources on concepts that are not ready for publication.

The final quality will still depend on the original footage, model output and upscaling process.

Where Can Users Access Gemini Omni 1.2 Flash?

Google announced general availability through the Gemini API and Gemini for business.

The company is also bringing the model into several of its creative and business products, including Google Vids, Flow and Google Ads.

Availability and access conditions can vary depending on the product, subscription and rollout schedule.

Not every platform necessarily offers identical features, limits or pricing.

Developers using the API should consult the applicable documentation and pricing information, while consumers should check their Gemini subscription and regional availability.

What Does the New Model Mean for Content Creators?

The lower pricing and longer video duration could be particularly useful for creators who regularly publish short-form content.

YouTube Shorts, Instagram Reels and TikTok videos often rely on short visual sequences, product demonstrations and promotional clips.

AI video generation can help creators produce supporting footage without recording every scene themselves.

For example, a technology website could use generated visuals to illustrate a new smartphone feature, while an online retailer could create short promotional videos for products.

Small businesses could also experiment with advertising concepts without immediately investing in conventional video-production resources.

However, AI-generated footage still requires human review. Product details, text, logos, hands, facial features and physical movements can sometimes contain errors.

Creators should also ensure that generated content does not misrepresent real products, people or events.

What Are the Limitations of Gemini Omni 1.2 Flash?

Despite the improvements, the new model is not a complete replacement for professional video production.

Google’s announcement highlights improvements in consistency and reference following, but generative video systems can still struggle with complicated scenes.

Fast sports movements, rapid script changes and complex physical interactions may produce inconsistent results.

Businesses should also consider the time required to review and correct generated footage. A lower generation price does not automatically translate into a lower total production cost if many attempts are needed.

Other considerations include:

• Accuracy of product appearance.
• Consistency between scenes.
• Dialogue and lip-sync quality.
• Compliance with copyright and privacy requirements.
• Subscription and API access restrictions.
• Additional costs associated with production workflows.

Google says generated videos include an imperceptible SynthID watermark, which is designed to help identify AI-generated content.

The company also applies restrictions to certain capabilities, including audio-reference functionality.

How Does Gemini Omni 1.2 Flash Compare With Omni 1.1 Flash?

The new version introduces improvements across price, video duration and editing controls.

The comparison shows that Google is focusing on both affordability and production flexibility.

The new model is not simply a cheaper version. Its longer clips and editing features could also reduce the number of separate generations needed for certain projects.

Final Word

Google’s Gemini Omni 1.2 Flash brings a 30% price reduction to AI video generation while increasing the maximum video duration to 15 seconds.

At $0.07 per second for 720p generation through the API, a 15-second clip costs approximately $1.05 before additional processing costs.

The model also introduces expanded conversational editing, improved reference consistency, more natural dialogue and options for drafting videos at lower resolutions before upscaling them.

For developers, advertisers, online retailers and content creators, these changes could make AI-powered video production more practical.

However, output quality, access restrictions and the need for human review remain important considerations. The real impact will depend on how reliably the model performs in everyday production workflows.