Gemini Omni

Overview

Gemini Omni Flash's image-to-video capability is aimed at developers who need to quickly turn a single still image into a deliverable short video. Just provide one image and a prompt describing the motion and sound of the scene, and the model outputs a finished clip with audio. Parameters let you choose a duration of 4 to 10 seconds, a resolution from 720p to 4K, and landscape or portrait framing. It suits teams producing ad assets, social media short videos, and concept demos, and is called through a unified API on HiAPI.ai with pay-per-use billing.

Key Highlights

• A single prompt describes both visuals and sound — the prompt describes how the input image should move and sound, so one generation already includes visual motion and audio, reducing the need for post-production voiceover.
• Output specs selectable by delivery channel — supports 4/6/8/10 second durations, 720p/1080p/4K resolutions, and 16:9 and 9:16 aspect ratios, fitting both landscape and portrait placements.
• Fixed seeds for reproducible results — a seed parameter lets the same input reliably reproduce the same result, which helps with debugging and comparing results during batch generation.

Capabilities

Video with sound from a single image: takes one input image and a prompt and outputs a video containing motion and audio, skipping the intermediate steps from storyboard to editing.

Parameterized duration and aspect ratio: set duration anywhere from 4 to 10 seconds and use aspect_ratio to choose landscape or portrait, producing assets in the specs your placement requires.

Tiered resolution output: supports three output tiers—720p, 1080p, and 4K—so you can validate quickly at low resolution during iteration and switch to high resolution for the final version.

Batch generation with controllable seeds: passing the same seed reproduces the same result, making it easy to lock in a style baseline across batch jobs and pinpoint differences caused by parameter changes.

Use Cases

• E-commerce operations: product images to short videos — upload a still product photo and describe the display motion to generate a 9:16 portrait video for feed ads and product detail pages.
• Social media creators: dynamic content from a single illustration — use an illustration or photo as input to generate a few seconds of dynamic footage as fresh material for short-video platforms, lowering the cost of frame-by-frame production.
• Design teams: motion previews of concept art — turn design drafts or concept art into short clips with sound to show clients or teammates the pacing and atmosphere before deciding whether to move into full production.

Limitations and Selection Advice

Currently each task accepts only one input image, prompt is required and must describe both visual motion and sound, output duration is limited to 4 to 10 seconds, and only 16:9 and 9:16 aspect ratios are supported—no multi-image reference or shot extension. This capability is not yet launched (comingSoon) and cannot be called for now. For scenarios needing multi-image reference or multimodal combined input, choose another route in the same family; actual availability is subject to the platform's launch status.