# Omnivideo GPT Image 2.5

> Canonical: https://omni-video.art/pl/gpt-image-2-5

Describe a single moment with a clear focal point. Compare alternate compositions as still images before deciding which direction to develop in video; a matching prompt does not guarantee continuity across separate generations.

## Omnivideo GPT Image 2.5 sequence concepts

### Will separate stills stay consistent?

No continuity guarantee is documented. Use references and stable descriptions to guide the appearance, then inspect each result before assembling a sequence.

## Technical specs

### 20,000

Prompt characters

### 30 MB

Per reference file

### 1K–4K

Output size options

### 1–4

Images per batch

## Who should use sequence studies?

### A practical starting point

Directors can compare a sequence of candidate visual moments, then select individual images for further development.

## Questions about sequence concepts

### What distinguishes the variants?

Flare and Sunburst have separate generation endpoints. Their published controls match, so compare actual images using the same input instead of assuming one is faster or more accurate.

### Which frames can use a larger output?

2K and 4K are available except at 8:9, 9:8, 16:27 and 27:16. Those four shapes require 1K; changing the ratio restores higher-size choices.

### Can I request empty space behind an object?

Opaque, Auto and Transparent backgrounds are supported. At higher resolutions Sunburst needs an isolated-subject prompt without backdrop or shadow; an editing request should also specify extraction.

## Continue with your image

## Powiązane narzędzia

### [Generator obrazów Omnivideo AI](https://omni-video.art/pl/ai-image-generator)

### [Omnivideo Gemini Omni — generator wideo AI](https://omni-video.art/pl/ai-video-generator)

_Ostatnia aktualizacja: 21 lipca 2026_