What is a GPT Image 2 generator?

It is an image generation workflow that creates or revises still images using GPT Image 2. In ShotSlate, the image prompt, references, output, ratio, resolution, and downstream video connections stay on the same project canvas.

Can ShotSlate generate images from text only?

Yes. ShotSlate supports text-to-image generation with a prompt, aspect ratio, and resolution setting.

Can I use reference images?

Yes. ShotSlate can pass connected image references into the image generation request. This is useful for visual direction, product cues, style references, and continuity between frames.

Which image sizes does ShotSlate support?

ShotSlate exposes seven aspect ratios and plan-based 1K, 2K, and 4K resolution options. The generated pixel size must also stay within the current GPT Image pixel limits enforced by the product.

Is this the same as a video generator?

No. This page is for generating still frames. If you want motion, use an approved generated image as a first frame in the Seedance image-to-video workflow.

gpt image 2 generator

GPT Image 2 generator for frames that move forward

Create a still frame from a prompt, guide revisions with references, choose aspect ratio and resolution, then keep the image connected to the video clip or composition it supports.

See the workflow

This guide is based on ShotSlate’s current GPT Image 2 image generation workflow. It covers text-to-image, reference-guided image generation, aspect ratio, resolution, and downstream canvas connections that are available in the product today.

Updated July 28, 2026

External verification

Model facts checked against public sources

These references are used for model context only. ShotSlate product claims on this page are limited to the controls available in the current implementation.

OpenAI GPT Image 2 model pageLists GPT Image 2 as an OpenAI image generation model with text and image input, image output, and support for image edits.OpenAI image generation guideDocuments the image generation and editing workflow used for prompt-based and reference-guided image creation.

Current product workflow

From prompt to approved frame on one canvas

GPT Image 2 is most useful in ShotSlate when an image is treated as a production frame, not a detached download. The image can hold visual direction, references, settings, and the next video step in the same project.

01
Write a frame brief
Describe one image: subject, composition, visual style, lighting, and the shot’s role in the sequence. Keep the prompt focused enough to review as a single frame.
Current image generation accepts prompts up to 5,000 characters and stores the displayed brief with the image node.
02
Add reference images when needed
Connect approved images as references when the frame needs a visual anchor, product cue, mood board, or prior shot context.
Current image generation supports text-to-image and image-to-image behavior with up to nine reference image URLs.
03
Send approved frames forward
Use the generated still as a storyboard frame, a first frame for Seedance video generation, or a source asset in a composition.
ShotSlate keeps image result nodes, prompt context, reference connections, generated media, and downstream video nodes on the same canvas.

Frame brief checklist

What to specify before generating with GPT Image 2

A good image prompt should describe the exact frame you want to review. The goal is not to write a whole film, but to give the model enough visual direction to create a usable production still.

Need to animate the frame?

Once a still frame is approved, connect it to an image-to-video workflow instead of asking text-to-video to reinvent the composition.

Open Seedance image to video

Brief partWhy it mattersExample

SubjectDefines the main person, product, prop, or environment that should anchor the image.A matte black smart speaker on a concrete kitchen counter

CompositionTells the model how the frame should be staged instead of leaving the camera angle random.Centered hero shot, 35mm lens feel, low three-quarter angle

LightingControls the mood and helps the generated image match the planned scene.Soft window light from the left with subtle cyan fill

StyleSets the creative direction without overloading the prompt with unrelated adjectives.Premium product photography, minimal, realistic texture

Reference roleExplains what connected images should influence when references are used.Use the reference only for product shape and color palette

Avoid listReduces common artifacts or unwanted details before credits are spent.No extra text, no distorted logo, no duplicate products

Workflow illustrations

Treat generated images as traceable production assets

These illustrations show how ShotSlate keeps image prompts, references, output frames, and downstream video steps connected. They demonstrate product behavior; they are not presented as independent customer work.

Step 01

Generate a shot frame

Create one still frame with prompt, ratio, and resolution settings visible beside the result.

Illustrated ShotSlate canvas with a GPT Image 2 generation node and generated frame.

Step 02

Use references with intent

Connect existing images as visual context, then keep the reference relationship visible for later review.

Illustrated ShotSlate canvas connecting reference images to a generated still frame.

Step 03

Move frames into motion

Send approved images into video nodes or compositions so the frame remains tied to the sequence.

Illustrated ShotSlate composition workflow using GPT Image 2 frames as source assets.

Production notes

GPT Image 2 controls currently available

These values reflect the current ShotSlate implementation and plan gates, not every possible capability of the upstream model.

StageSystemAvailable controls

Input modeShotSlate image nodeText-to-image · image-to-image with connected references

Reference inputsGPT Image 2 requestUp to 9 reference image URLs

Image settingsShotSlate generation controls7 aspect ratios · 1K, 2K, and 4K by plan

Limits to plan around

What this page does not promise

A generated image still needs review

GPT Image 2 can create useful frames, but creators should still check composition, artifacts, brand details, and whether the frame fits the sequence.

References guide the result; they do not guarantee cloning

Reference images can influence the output, but exact product markings, faces, typography, and fine details may still change and should be verified manually.

Resolution depends on plan access

ShotSlate exposes 1K on the free plan, 1K/2K on Basic, and 1K/2K/4K on Pro, Studio, and Scale plans.

Common questions

GPT Image 2 generator FAQ

Answers reflect the current ShotSlate implementation for GPT Image 2 image generation.

ShotSlate Product Team

Reviewed against OpenAI public GPT Image 2 documentation and the current ShotSlate implementation for image model ID, upstream model ID, prompt validation, reference image inputs, aspect ratios, image size limits, and plan-based resolution gates.

About ShotSlate Report an issue

Create a frame you can actually use downstream

Generate a still image, review the result, keep its references visible, and connect approved frames into video or composition steps.

What to specify before generating with GPT Image 2

A good image prompt should describe the exact frame you want to review. The goal is not to write a whole film, but to give the model enough visual direction to create a usable production still.