A Guth Labs publication ↗
Back to News

Gemini Omni 1.2 Flash lowers video pricing, with workflow limits worth reading

A curving strip of cyan glass cinema frames shows a copper paper bird in flight, with one frame outlined in amber to suggest editing.
AI-generated editorial illustration by guthlabs.newsroom

Google made Gemini Omni 1.2 Flash generally available on November 5, extending native video generation to 15 seconds and lowering the advertised 720p price to $0.07 per output second. Google describes that as a 30% reduction from Omni 1.1. The update also supports conversational revisions and improved reference following. Google launch announcement

At that output rate, one 15-second 720p generation is $1.05. That calculation is a starting point for a budget, not the full cost of a finished video: input charges, revisions, rejected attempts and any higher-resolution output can change the total.

The 15-second headline has exceptions

Google Cloud’s current model documentation specifies generation and editing durations from three to 15 seconds, but limits first-and-last-frame generation to three to ten seconds. Extensions add three to ten seconds under that documented route. Teams designing a workflow around exact durations should read those mode-specific limits before switching models. Google Cloud model specifications

The same page supports the Interactions API but not Chat Completions or GenerateContent. It also lists function calling and structured output as unsupported. An integration built around a text model’s interface therefore needs more than a replacement model name. Supported APIs and capabilities

For engineering teams, these are testable requirements. Confirm the requested mode, duration and resolution, then verify that the returned asset matches them. Preserve the generation settings with the output so a later revision does not depend on reconstructing the original request.

Provider pricing includes more than output seconds

Runway’s November 9 integration guide lists the model at $0.02 per second for 360p, $0.07 for 720p, $0.10 for 1080p and $0.21 for 4K. It also charges $0.01 per reference image and $0.01 per input-video second. Those figures describe Runway’s service and should not be assumed to be every provider’s billing policy. Runway API and pricing guide

Runway further states that input-moderation failures are charged and that generated download links expire. Cost controls and prompt review should therefore sit alongside a step that saves successful outputs. Runway billing and output guidance

Audio references and generated sound are different features

Google Cloud’s launch post says audio references require allowlisting. It separately advertises generated dialogue and improved lip synchronization. A buyer should not infer unrestricted voice-reference input merely because a model can produce sound. Google Cloud availability announcement

Cloud customers can also obtain provisioned throughput, and Google says outputs carry SynthID watermarks. The company still identifies fast sports action, rapid script changes and complex physics as areas under development. Cloud deployment options, Google’s stated limitations

For a production comparison, measure cost per accepted clip and time spent revising it. Check continuity, spoken content, product details and the final export rather than judging only a selected demonstration. Omni 1.2 Flash provides longer clips and a lower starting price; the useful saving is the one that remains after the clip passes those checks.

Sources and citations

Revision history

Published .