Gemini Omni 1.2 Flash: what it is, pricing, how to use it

Gemini Omni 1.2 Flash is Google's new video generation and editing model, and it makes clips with synced audio up to 15 seconds long at $0.07 per second for 720p. You can try it free in Google Vids, up to three 15-second videos a month, or pay per second through the Gemini API.

The posts on X

@GoogleAI announced it on November 5, saying Google "bumped the video generation and editing limit up to 15 seconds." The cap on Omni 1.1 was 10. The post also mentions better character speech and fewer visual glitches such as jump cuts.

@_philschmid gave developers the short version: "pricing reduced by 30% to $0.07/s for 720p," live in the Gemini API and AI Studio.

@GeminiApp added the plan detail. The model is rolling out "to all Google AI subscribers," but 15-second clips in the app are for Ultra subscribers.

What it costs

There is no free tier in the API. Google's pricing page bills by output tokens and lists these effective rates. The 15-second column is my multiplication, apart from the 720p figure, which is Google's.

The launch post from @Google says "Draft at 360p or 720p ($0.07/s)," which reads as if both cost the same. The pricing page shows 360p is cheaper, so drafting there saves about two-thirds.

Compare that with Pika H3 Flex, which also tops out at 15 seconds and charges $0.01 to $0.16 per output second at its launch prices, depending on resolution.

How to try it

The no-code route is Google Vids. Anyone with a Google account or Workspace subscription gets the three free clips a month.

Developers need a paid Gemini API key. The model runs on the Interactions API, not generateContent. This is adapted from Google's Omni docs, and I haven't run it myself:

import base64
from google import genai

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-omni-1.2-flash",
    input="A single unbroken tracking shot following a paper airplane gliding through a bustling library.",
    response_format={"type": "video", "duration": "15s", "resolution": "360p"},
)
with open("clip.mp4", "wb") as f:
    f.write(base64.b64decode(interaction.output_video.data))

To change that clip, pass its interaction ID as previous_interaction_id in a new request and describe the edit. Google's suggested edits include swapping a product, changing the lighting or changing the camera angle. Don't set store to false on a clip you may want to edit, because that blocks later turns.

Google's recommended flow is to draft at 360p or 720p, fix details with follow-up edits, then upscale the keeper to 1080p or 4K. Test one upscale and check your bill before you budget around it.

Limits worth knowing

These come from Google's docs:

• Clips run 3 to 15 seconds in whole seconds. If you don't set a length, you get 10 seconds, so a 720p clip costs about $0.70 by default.
• Aspect ratio is 16:9 or 9:16, so there's no square option.
• Extending a video adds 3 to 10 seconds per turn, up to 30 seconds total.
• First and last frame interpolation stops at 10 seconds.
• Clips of 11 to 15 seconds accept one reference video, while shorter clips accept up to three. Either way you get up to six reference images.
• Editing or extending uploaded video isn't available in the EEA, Switzerland or the UK.

If you already use Omni 1.1, you don't have to change anything. On December 3, API requests to gemini-omni-1.1-flash start routing to 1.2 at the lower rate. If you've tuned prompts for 1.1, retest them before that date.

My read

The price cut matters more than the extra five seconds. Video work is mostly retries, and Google's approach here is to let you fix a clip instead of paying to regenerate it. If the edits hold up, a finished clip could get cheaper by more than the 30% headline, but I'd test that before believing it. @DesignArena ran the kind of check I'd like to see more of, giving 1.1 and 1.2 the same prompt for a flight through the Colosseum. For your own work, skip the demo reels and count how many tries it takes to get a usable clip. Our FLUX 3 post covers why vendor-run video tests deserve some skepticism.

What this means for a small business

A 15-second vertical clip is about the length of a short social ad, and it now costs about a dollar at 720p. A service company could draft a seasonal promo from its own product photos before paying anyone to film it. All API output carries an invisible SynthID watermark, and you'd still want a person to review every clip before it runs.

The harder part is usually the process around the clip, like who writes the brief and who signs off on it. New Face Design's free process audit can help you map that work and see which pieces are worth automating.