Google launches Gemini Omni 1.1 Flash with 4K video and cheaper 360p previews

Google announced the Gemini Omni 1.1 Flash on August 27, 2026, an update to its video generation model aimed at developers. According to Google DeepMind product managers Anish Nangia and Alisa Fortin, this version's focus isn't a new quality baseline, but rather control and cost: features that make the model "production-ready" for professional use via the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform.

For those building software in Brazil who had been viewing AI video generation as too expensive and unpredictable for production, this set of updates addresses exactly the two variables that hold back a project: predictability of results and the end-of-month bill.

What Changed in Practice

The highlights announced by Google:

• Scene extension. The model now analyzes up to 10 seconds of prior context to continue a video, compared to just the last second in previous models. It's possible to extend in 10-second increments up to a cumulative total of 40 seconds, which Google describes as improving visual consistency and narrative adherence.
• First and last frame. You can specify the starting and ending keyframes of a take, and Omni 1.1 generates the continuous video between them, which the company points to as useful for camera orbits, zoom transitions, and loops.
• 360p drafts. Lightweight 360p previews run, according to Google, up to 60% faster and at a third of the cost compared to Omni 1.1's own standard 720p resolution. The speed figure comes with an asterisk: it's based on system throughput of 360p vs. 720p.
• Upscale up to 4K. Final outputs can be generated in 1080p or 4K for production.
• Video references in multimodal input. It's possible to reference up to three seconds of video to maintain visual context and character consistency.

The workflow pattern Google suggests is clear in these features: iterate cheaply at 360p until getting the idea right, and only then spend on high resolution. In the documentation, the call to extend a scene looks like this, already requesting the preview resolution:

from google import genai

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-omni-1.1-flash",
    previous_interaction_id=previous_video_interaction.id,
    input=[
        {"type": "text", "text": "Continue the scene."}
    ],
    response_format={
        "resolution": "360p",
    },
)

Cost Is the Central Argument

The point that most interests those putting this into production is the resolution-scaling model. By separating the cheap draft (360p) from the expensive delivery (4K), Google addresses the classic video generation problem: creators tend to generate several versions before landing on the right one. One of the application examples cited in the announcement, the "Draft Room," proposes exactly this: generating 3 to 4 variations at 360p, varying one element at a time and comparing them side by side.

For Brazilian teams that were already experimenting with these models, the practical effect is the ability to better estimate the cost per delivered video: the exploration phase, which is where most calls are burned, now costs a fraction. It's worth following the official Omni 1.1 Flash pricing table on Google AI Studio, since dollar values and the exchange rate still weigh on the bill for any operation here.

Who's Already Using It

Google listed customers already running Omni Flash in production via the Agent Platform API. Adobe integrated the model into Adobe Firefly, and Figma Weave and Runway also appear among those who adopted it. Figma's quote sums up the "control" argument that Google is selling in this version:

"Gemini Omni Flash is one of the strongest video models available in Figma Weave, where the canvas helps creative teams build on every generation, attaching references, branching different versions, and shaping something unique. With extensions, richer reference material, and 4K resolution, Gemini Omni Flash takes teams beyond generating videos to truly directing them."

>

-- Itay Schiff, Creative Director, Figma Weave

GMI Cloud, also cited, highlighted the model's precision for educational and explainer content, "where getting it right is essential."

The Community Pumps the Brakes

In the Hacker News thread about the launch, the reception mixes technical interest and skepticism. Developer Simon Willison noted the strategic contrast between the major players:

"Interesting that OpenAI abandoned Sora entirely but Google are continuing to invest heavily in their own video generation. Maybe because they see video generation as key to developing \"world models\"?"

>

-- simonw

The question of training data also came up. User Gecko4072 speculated about the origin of the material feeding these models: "So Seedance is good primarily because of TikTok and this because of YouTube. I wonder what portion of all recorded video is privately held in hard drives at people's homes or Apple photos."

There are still those unconvinced by the quality. Commenter polytely said they continue to experience "major uncanny valley from any of the videos featuring humans," saying that something about these videos still bothers them, a reminder that, despite the leap in control, the barrier of human realism remains visible to part of the audience.

What Remains Open

The announcement doesn't detail the full pricing table in text (it appears as an image in the original post), so fine-grained calculation of cost per token or per second of video depends on consulting Google AI Studio directly. There's also no independent benchmark comparing Omni 1.1 with competitors; the only figures released (60% faster, a third of the cost) are Google's internal measurements against its own 720p version. For teams evaluating adoption, the sensible path is to test the 360p-to-4K workflow on a real case before assuming the announced gains are guaranteed.