Can GPT Image 2 start from text and existing images?

Yes. The Toono AI workspace supports both text-to-image and reference-led image workflows for GPT Image 2. Use references when identity, structure, or material needs to carry into the new result.

How many references and outputs can I use here?

The current workspace accepts up to 10 reference images and can request 1 to 4 outputs. The interface shows the active limits, and larger requests generally use more credits and take longer.

Which size, quality, and file formats are available?

The current setup offers square, landscape, and portrait sizes; low, medium, or high quality; and PNG, JPEG, or WebP output. Choose the final placement before increasing quality.

Is generated text ready to publish without checking?

No. Even when short text looks convincing, verify every character, number, line break, and legal phrase. For critical or frequently changing copy, generate the visual structure first and add final typesetting in a controlled design tool.

OpenAI image creation for production briefs

Build a connected visual set from one brief

Start from text, add references when identity or structure matters, and produce campaign layouts, catalog views, diagrams, or edits from one GPT Image 2 workspace. This page explains the workflow; production controls open in the dashboard.

Create with GPT Image 2

Open image generation with this model selected

See examples and review points

Text in images
References
Image editing

One editorial system connects a text-led poster, product catalog views, a three-step guide, and the same lamp across four environments.

Visual examples

Four production tasks, not four versions of the same picture

A model page should show where the model earns its place in a real workflow. These examples focus on readable short text, detailed product presentation, structured information, and reference continuity.

OpenAI

Design a text-led campaign layout

Pair a concise headline with a clear hierarchy, deliberate negative space, and a visual system that can move into a launch page or event campaign.

OpenAI

Explain a process as a visual guide

Turn a short sequence into a legible diagram with distinct steps, consistent symbols, and enough space for editorial use.

OpenAI

Present one product from useful views

Show form, finish, open and closed states, and a material close-up so a catalog team can evaluate more than a hero angle.

OpenAI

Carry a reference subject across scenes

Compare silhouette, color, hardware, and proportions when the same object moves from a desk to a living room, studio, or campaign set.

Examples demonstrate creative directions, not guaranteed outputs. Results, latency, credit cost, available sizes, and reference fidelity depend on the chosen settings and source material. Review every asset before publishing.

Practical workflow

Build the asset set from constraints, not prompt decoration

GPT Image 2 responds best to a production brief that separates what must stay fixed from what can change between outputs.

01
Define the placement and non-negotiables
Name the final channel, required text, subject details, brand colors, composition limits, and any elements that must not appear.
02
Add references with a specific job
Use each reference for a clear purpose such as subject identity, palette, layout, or material. Fewer well-chosen references are easier to evaluate than an unstructured mood board.
03
Choose output settings, compare, then refine
Set size, quality, format, and output count for the placement. Select the strongest direction and revise one structural issue at a time.