Gemini Omni 1.1 Flash Video Extend API by Google

Gemini Omni 1.1 Flash — Video Extend

Model ID: google/gemini-omni-1.1-flash/video-extend

Gemini Omni 1.1 Flash is Google DeepMind's natively multimodal model for video generation and editing. This variant — new in 1.1 — takes an existing clip and continues it, appending a seamlessly matched 3-to-10-second continuation. Chain the calls and a single shot grows to a total of up to 40 seconds of coherent video.

Overview

Gemini Omni Flash is a single transformer that processes text, images, audio, and video simultaneously rather than stitching separate modality-specific pipelines together. The result is output that is more cohesive, more consistent across turns, and more controllable than a chained generate-then-dub workflow.

What sets it apart from earlier video models is that it natively generates audio with every video — dialogue, ambience, music, and sound design are produced together with the picture rather than added afterward. The model is grounded in Gemini's real-world knowledge, so it reasons about physics, narrative logic, culture, and visual composition to produce results that feel intentional and cinematic. Every generated clip carries an imperceptible SynthID watermark alongside C2PA Content Credentials for provenance verification.

For extension specifically, the model reads up to the last 10 seconds of the source as context and generates what comes next — carrying motion, lighting, subject identity, and the soundtrack across the join so the continuation reads as one take rather than a cut.

What's new in 1.1

Version 1.1 is a control-and-fidelity upgrade over the first Omni Flash release:

• Flexible resolution output — new 4K and 1080p rendering for finished work, plus a fast 360p draft mode for previewing a shot before committing to a full-quality render.
• Longer continuous generation — scene extension analyses up to 10 seconds of prior context and grows a shot in increments, to a total of up to 40 seconds. This is the endpoint that exposes it.
• Precise shot start and end control — first and last frame can both be supplied, producing smoother camera moves, scene transitions, and seamless loops.
• Character and style consistency — a new video reference capability (clips of up to 3 seconds) substantially improves subject and art-direction stability.

How the five endpoints relate

AtlasCloud exposes Gemini Omni 1.1 Flash through five model IDs — text-to-video, image-to-video, reference-to-video, video-edit, and video-extend. All five route to the same underlying gemini-omni-1.1-flash model and differ only by the input modality they accept, corresponding to the model's task parameter (text_to_video, image_to_video, reference_to_video, edit, extend). This endpoint maps to extend.

Inputs

This variant takes a source video and a text prompt describing what happens next, with optional reference images.

• Prompt — natural-language description of how the source video should continue, up to 20,000 characters.
• video (required) — the source clip to extend. Must be a publicly accessible MP4 URL, 10 seconds or less; the model uses up to the last 10 seconds as context for the continuation.
• reference_images (optional) — 1 to 10 images used as character, scene, or style references. PNG, JPEG, JPG, or WebP, each up to 20 MB, supplied as public URLs or base64-encoded images. Useful for introducing something new into the continuation while keeping it on-model.

Referencing images from inside the prompt

Instead of describing an uploaded image in words, point at it directly with <IMAGE_REF_N> — the Nth entry in reference_images, 0-based, so the first image is <IMAGE_REF_0>. Bindings may also be declared up front with a [# References <TAG>@MediaN] block at the start of the prompt, where the @MediaN handles are 1-based (@Image1 is the same item as <IMAGE_REF_0>).

Example: The camera keeps pushing in as the character from <IMAGE_REF_0> steps into frame from the right.

Key Capabilities

• Seamless continuation — append 3 to 10 seconds that match the source's motion, lighting, subject, and audio across the join.
• Long-form assembly — chain extensions to build a single coherent shot of up to 40 seconds total.
• Context-aware generation — up to the last 10 seconds of the source informs what comes next.
• Reference-guided continuation — introduce a subject, object, or style into the new segment with up to 10 reference images, addressed inline with <IMAGE_REF_N>.
• Resolution control on the continuation — render the appended segment at 360p draft quality or up to 4K.
• Native audio generation — the continuation carries its own synchronized soundtrack.
• Adjustable reasoning — the thinking_level control trades latency for quality on complex continuations.
• Reproducible results — set a fixed seed to reproduce or iterate on a specific generation.

Input Parameters

duration describes only the appended segment, not the finished video. Aspect ratio follows the source, so this endpoint has no aspect_ratio parameter.

Prompting Tips

• Describe what happens next, not the scene as a whole. The model already has the source as context; the prompt only needs to carry the story forward.
• Keep continuity cues explicit when they matter — "the camera keeps moving left," "the music stays on the same beat" — so the join is invisible.
• Extend in increments and review each pass. Chaining several deliberate 5–10 second continuations gives more control than one long request.
• Draft the continuation at 360p to check the beat lands, then re-run with the same seed at full resolution.
• Extending uploaded videos is unavailable in the EEA, Switzerland, and the UK.

Use Cases

• Long-form shots — grow a 10-second generation into a 40-second sequence without a visible cut.
• Narrative continuation — carry a scene forward beat by beat while keeping characters and setting stable.
• Ad and social cutdowns — build a longer master from a short hero clip, then trim to length.
• Loop and filler extension — lengthen background plates and ambient footage to fit a timeline.
• Storyboard expansion — take an approved opening shot and continue it into the next action.

Pricing

Billing combines a per-second charge for the generated continuation at the requested resolution, a per-reference-image charge, and a per-second ingestion charge for the source video whose duration is parsed from the supplied file.

SKUs

Formula

source_duration = duration parsed from the source video

cost = min(10, max(3, duration)) × resolution_rate
     + count(reference_images) × $0.00168
     + source_duration × $0.0087

Where duration is the requested length of the continuation and resolution_rate is $0.041 for 360p, $0.11 for 720p, $0.16 for 1080p, and $0.31 for 4k.

Notes and examples

• The generation term covers only the appended segment and is clamped to 3–10 seconds: a shorter request is billed as 3 seconds, a longer one as 10.
• The ingestion term is charged on the measured duration of the source video you pass in, at $0.0087 per second — so each link in an extension chain pays for the clip it continues from.
• Reference images are billed per item, not per second.
• duration defaults to 10 and resolution defaults to 720p when not supplied.
• A 10-second 720p continuation of a 10-second source, no references, costs 10 × $0.11 + 10 × $0.0087 = $1.187.
• A 5-second 1080p continuation of an 8-second source with 1 reference image costs 5 × $0.16 + 1 × $0.00168 + 8 × $0.0087 = $0.87128.
• A 3-second 360p draft continuation of a 10-second source costs 3 × $0.041 + 10 × $0.0087 = $0.21.
• A 10-second 4K continuation of a 10-second source with 2 reference images costs 10 × $0.31 + 2 × $0.00168 + 10 × $0.0087 = $3.19036.