Gemini Omni 1.1 Flash: API, Pricing, 4K Video, and Complete Guide

Gemini Omni 1.1 Flash is Google's production-ready model for fast video generation, conversational editing, interpolation, and scene extension. Released as the stable Gemini API model gemini-omni-1.1-flash on August 27, 2026, it adds the controls that were missing from the earlier preview: 360p drafts, 1080p and 4K upscaling, first-and-last-frame generation, and longer stories built through video extension.

On Poyo.ai, the new model starts at 45 credits ($0.225) for a four-second generation without video input. It is a separate model from the earlier omni-flash integration; the model ID, prices, and capabilities in this guide apply only to gemini-omni-1.1-flash.

Status checked August 31, 2026. Google is actively expanding Gemini Omni. Confirm current request fields, regional availability, safety limits, and pricing before shipping a production workflow.

Gemini Omni 1.1 Flash at a glance

What is Gemini Omni 1.1 Flash?

Gemini Omni 1.1 Flash is a multimodal video model designed to generate and modify video through instructions rather than a traditional editing timeline. You can begin with text, an image, an existing video, or a combination of reference media, then describe the result you want.

Its most important difference from a conventional text-to-video model is conversational editing. A first interaction can create a clip. A follow-up interaction can change lighting, replace an object, alter the background, or continue the scene while preserving the previous video state. The Gemini Interactions API tracks that state through a previous_interaction_id, so an application does not need to restate the full creative brief on every turn.

Google positions the model around three ideas:

• Native multimodality: video generation can use text, image, video, and other media context together.
• Conversational control: creators can refine an existing result with natural-language instructions.
• World knowledge: the model uses Gemini's broader understanding of objects, places, physics, and cultural context when constructing a scene.

These strengths do not remove the need for testing. Fast body movement, precise hand-object contact, long identity consistency, and complicated physical interactions can still fail.

What changed in Gemini Omni 1.1 Flash?

The 1.1 release turns the original preview into a stable developer model and adds controls aimed at production workflows.

Scene extension with more context

Gemini Omni 1.1 Flash can continue an existing video from its final frames. Google says the model can analyze up to ten seconds of prior video context, rather than relying only on the final second. Extensions can be generated in ten-second increments up to a cumulative length of 40 seconds in Google's documented workflow.

This is useful for building a longer sequence from short approved segments. It does not guarantee perfect continuity. Check faces, clothing, props, lighting direction, camera speed, and background geography at every join.

First and last frame interpolation

Provide a starting image and an ending image, then describe how the camera and subjects should move between them. The model creates the transition rather than treating one image as a loose visual reference.

This control is especially useful for:

• camera orbits and push-ins;
• product reveal transitions;
• seamless loops;
• location changes;
• matching a shot to an existing edit point;
• holding a character's look across the beginning and end of a clip.

Faster 360p drafts

Google introduced 360p for lightweight previews. According to the launch announcement, 360p can be up to 60% faster and cost roughly one third as much as standard 720p generation in Google's workflow.

The practical pattern is simple: explore prompts and composition at 360p, select a promising result, then render or upscale the approved version at a higher resolution. Teams should still confirm that motion, small text, faces, and fine objects survive the higher-resolution step.

1080p and 4K upscaling

The API accepts 1080p and 4k resolution choices. Google describes both as upscaled outputs. They are intended for delivery and post-production, but a larger frame does not repair an incorrect pose, a continuity error, or weak prompt adherence. Approve content and motion before paying for the final resolution.

Video references

Omni 1.1 can use short video references alongside images and text. Google's launch examples use reference performances to guide different characters in one scene. Reference video can carry motion, timing, or visual context that is difficult to express with text alone.

Gemini Omni 1.1 Flash pricing on Poyo.ai

Poyo.ai uses 200 credits = $1. Pricing depends on whether the request includes a video and whether the output is 4K.

No video input

The same Poyo.ai tier applies to 360p, 720p, and 1080p for requests without a video input. That makes 360p useful for fast exploration, while 1080p provides better delivery value when the prompt is already approved.

With video input

Video-input jobs use a fixed generation price instead of the duration table above.

How Poyo.ai compares with Google's published rate

Google's Standard Gemini API pricing lists an effective rate of approximately $0.10 per second for 720p video output. On an effective per-second basis, Poyo.ai's no-video-input 720p tiers are approximately 44% to 55% lower, depending on duration.

This comparison applies only to the 720p, no-video-input scenario. Google's accounting for 360p, upscaled 1080p/4K, and editing workloads is not identical, so those modes should not be assigned the same discount claim.

Choosing the right resolution and duration

Do not use resolution as a substitute for art direction. A carefully planned 720p shot is usually more valuable than an inconsistent 4K shot.

How to use Gemini Omni 1.1 Flash

1. Define the task before selecting media

Choose one primary job: generate, animate, interpolate, edit, or extend. A request that tries to change the character, location, camera, action, art style, and timing at once is difficult to evaluate and difficult to repair.

2. Prepare clean reference material

For image-guided work, use references that clearly show the subject, clothing, product, or environment you need to preserve. For video editing, begin with a short source clip without unnecessary cuts. Google documents up to ten seconds of video context for editing and extension.

3. Describe scene, camera, time, and audio

A useful production prompt separates four things:

Scene: [subject, environment, lighting, important objects]
Camera: [shot size, lens feel, movement, whether cuts are allowed]
Timing: [what happens during each time range]
Audio: [dialogue, ambience, sound effects, or no dialogue]

For a single continuous shot, say so explicitly. Omni may otherwise create several cuts to make the clip feel more like a short narrative.

4. Draft before the final render

Use a short duration and low resolution to validate composition, motion, and prompt adherence. If the result is structurally wrong, do not upscale it. Change one instruction at a time so you can identify what improved or broke the shot.

5. Review continuity frame by frame

Check:

• face and body identity;
• hands and object contact;
• clothing and accessories;
• readable text;
• shadows and reflections;
• background layout;
• lip and audio synchronization;
• the join between an original clip and an extension.

Prompt patterns that work well

Single unbroken scene

In a single continuous shot with no cuts, a small brass automaton crosses a rain-wet workshop table. The camera tracks beside it at eye level. Warm window light reflects in shallow puddles. Quiet mechanical clicks and distant rain, no dialogue.

Timecoded product reveal

[0-2s] Macro close-up of a sealed glass bottle on a dark stone pedestal.
[2-5s] The camera slowly orbits as condensation forms and catches warm side light.
[5-8s] The background opens into a clean coastal sunrise while the bottle remains unchanged.
One continuous shot, no text, restrained ambient sound.

First-and-last-frame transition

Move smoothly from the supplied first frame to the supplied last frame. Use a slow clockwise camera orbit and preserve the character's face, clothing, and proportions. The transition must be continuous with no jump cut or new character.

Conversational edit

Keep the person, camera movement, timing, and room unchanged. Replace only the red chair with a weathered wooden bench. Match the original lighting, shadows, reflections, and contact points.

Scene extension

Continue the scene from the existing camera path. The character reaches the open doorway, pauses, and looks outside as morning light enters the room. Preserve the face, coat, room layout, movement speed, and ambient audio.

Migrating from Gemini Omni Flash Preview

Google lists gemini-omni-flash-preview for shutdown on September 30, 2026, with gemini-omni-1.1-flash as the replacement.

Migration should include more than changing a string:

1. Replace the preview model ID in every environment.
2. Test each input mode separately.
3. Add validation for 360p, 720p, 1080p, and 4K.
4. Test first-and-last-frame requests with exactly two ordered images.
5. Recheck video-upload and extension limits.
6. Compare output audio, duration, safety responses, and file-delivery behavior.
7. Recalculate budgets using the stable model's pricing.
8. Remove preview-specific fallback logic before the shutdown date.

Keep a small regression set containing faces, products, fast motion, readable text, dialogue, and extension seams. A successful HTTP response does not prove that the creative behavior remained unchanged.

Gemini Omni 1.1 Flash vs other video models

Choose with representative work, not one launch clip. Test the same prompt, references, target duration, and delivery resolution. Measure cost per approved clip, not cost per request.

Limitations and responsible use

Gemini Omni 1.1 Flash is more controllable than the preview, but it is still generative video.

• Fast movement can produce body-physics errors.
• Hands, tools, and contact between objects may break.
• A character can drift across long or heavily edited sequences.
• Upscaling improves delivery resolution, not semantic correctness.
• Video editing availability can vary by region.
• English is fully supported; Google says other languages have not been evaluated to the same degree.
• Safety filters apply to prompts, reference media, and generated results.
• Generated videos contain SynthID for provenance.

Obtain permission for people, voices, performances, brands, and source footage. Do not use conversational editing to impersonate a real person or misrepresent an event.

Frequently asked questions

What is the Gemini Omni 1.1 Flash model ID?

The stable Gemini API model ID is gemini-omni-1.1-flash. It is distinct from both Google's gemini-omni-flash-preview and Poyo.ai's earlier omni-flash model.

Does Gemini Omni 1.1 Flash support 4K?

Yes. The API supports 360p, 720p, 1080p, and 4K choices. Google describes 1080p and 4K as upscaled outputs.

How long can Gemini Omni 1.1 Flash videos be?

The model specification lists 3-10 seconds per generated output. Google's extension workflow can build longer stories in increments, up to a documented cumulative length of 40 seconds.

How much does Gemini Omni 1.1 Flash cost on Poyo.ai?

Without a video input, prices start at 45 credits ($0.225) for four seconds at 360p, 720p, or 1080p. Four-second 4K starts at 105 credits ($0.525). Video-input generation costs 120 credits ($0.60) for 360p/720p/1080p or 180 credits ($0.90) for 4K.

Can it edit an existing video?

Yes. The model can accept uploaded video and natural-language edit instructions. Regional availability and input limits should be checked before production use.

Can it use first and last frames?

Yes. Supply two ordered images and describe the transition. The model interpolates a video between them.

Is 4K generated natively?

Google documents 1080p and 4K as upscaled outputs. Review the underlying motion and details before treating an upscale as a final asset.

What happens to the preview model?

Google lists gemini-omni-flash-preview for shutdown on September 30, 2026 and recommends gemini-omni-1.1-flash as its replacement.

Build with Gemini Omni 1.1 Flash on Poyo.ai

Gemini Omni 1.1 Flash is most useful when generation and editing are treated as an iterative production system: draft cheaply, approve motion, refine one change at a time, and pay for 4K only when the scene is ready.

Open the Gemini Omni 1.1 Flash model page to try the model, review the current request schema and pricing table, and create an API key.

Sources

• Google: Build with Gemini Omni 1.1 Flash
• Gemini API model specification
• Gemini Omni generation and editing guide
• Gemini Developer API pricing
• Gemini API deprecations