Google cuts Gemini Omni Flash 1.2 video pricing to $6.26 per minute at 1080p

If your product creates ads or videos for product listings, a reduction in generation pricing feeds directly into your production costs. According to Artificial Analysis, a minute of 1080p video with Gemini Omni Flash 1.2 costs $6.26—roughly 31% less than with version 1.1. For a team running thousands of generations, that is a reason to recalculate the budget and check how much it costs to produce a video ready to deliver to a client.

Take a simple example: a thousand generated minutes at the new price would cost about $6,260. But a thousand minutes of finished content may require far more generation attempts. So the useful question for a developer is: how many attempts will the new model need to complete our task?

What exactly got cheaper

The original Artificial Analysis post says Google reduced the Gemini API rate from $17.50 to $12 per million output video tokens. This is the billing unit for generated video. The $6.26 estimate translates it into a figure that is easier to use for planning: the cost per minute at 1080p resolution.

The Artificial Analysis leaderboard lists the previous version at $9.12 per minute. The difference is $2.86, yielding savings of about $2,860 across a thousand generated minutes. This is a calculation based on published prices, not a forecast of your application's total bill.

It is easy to confuse two different tasks here. These figures concern video creation, so they cannot be used to estimate the cost of analyzing recordings uploaded by users. A service that watches hour-long interviews and produces summaries needs a separate calculation using the relevant model's rates.

Why a cheaper minute does not guarantee savings

Imagine a service that creates ten-second product videos. Scaling the published price linearly, each attempt costs about $1.04. If it takes three attempts to get a suitable result, generation costs per accepted video rise to about $3.13. This example excludes editing and human review.

This is a hypothetical scenario, not a measurement of Gemini's quality. It shows why comparing models solely on their price per second can put a budget at risk. A model may be cheaper but more frequently change the shape of packaging, get scene details wrong, or produce motion the client rejects. Some of the discount then gets spent on retries.

In its version 1.2 announcement, Google claims more faithful adherence to visual and audio references. The company has also increased the maximum length of generated and edited clips from 10 to 15 seconds. These suggest concrete hypotheses for product testing: does the product's appearance remain consistent, and can the model produce the required scene in fewer attempts? The vendor's claim alone does not answer those questions.

What to check before switching

Start with a small selection of real tasks: a product close-up, a person in motion, several objects in the frame. Use the same source materials and requirements for both versions. Define in advance which errors make a video unusable; otherwise, an impressive image can easily outweigh a failure to follow the brief.

For each scene, record the number of attempts needed to reach an accepted result, total costs, and waiting time. Separately, assess how much work remains for a person. If someone needs to spend a long time fixing details or rebuilding the scene after generation, that also affects the product's economics.

For a team already using Omni Flash 1.1, the price reduction is reason enough to run this comparison. For a new service, it adds another candidate to test. It is too early to conclude from the discount that this is the best-value offering on the market: results depend on resolution, scene requirements, and the proportion of successful takes. The practical benefit will come where the cost per accepted video falls.