Image Generation Module

Create and refine scientific figures from text, sketches, references, and existing drafts.

What This Module Does

Image Generation is where you create the figure. It helps turn a scientific idea into a visual draft, then gives you ways to revise that draft without starting over every time.

Use this module when you want to:

create a figure from a written description;
use a sketch or reference image to guide layout and style;
improve an existing figure draft;
correct labels or terminology;
redraw one confusing area;
make the image clearer before export or vectorization.

Starting Points

Text to Figure

Describe the scientific story, layout, labels, and intended use. This is the fastest way to create a new graphical abstract, mechanism diagram, or workflow.

Sketch to Figure

Use a rough drawing when the structure matters more than visual polish. A sketch can communicate panel order, arrows, and object placement.

Reference to Figure

Add examples, screenshots, or previous drafts when you want to preserve a visual direction, level of detail, or composition.

Figure Refinement

Start from an existing draft and ask FigRay to simplify, clean, relabel, or clarify it.

Best Fit

Figure type	What to provide
Graphical abstract	Study summary, desired stage order, main labels.
Mechanism diagram	Key actors, interactions, arrows, labels, and direction.
Protocol schematic	Step list, order, grouping, and output format.
Slide figure	Main message, audience, and desired level of detail.
Existing draft cleanup	Current image plus what should stay and what should change.

Basic Workflow

Describe the figure goal in one or two sentences.
Specify layout, panels, flow direction, and required labels.
Add sketches or reference images when visual guidance matters.
Generate the first draft.
Review scientific accuracy, composition, and text.
Use follow-up instructions for global revisions.
Use focused edits for labels, local regions, clutter, and clarity.
Export directly, or move to Vectorization when the figure is close to final.

The workspace currently provides GPT Image 2 and Nano Banana 2. Models can respond differently to composition, text, style, and revision instructions. If the first result is weak, compare models while keeping the request consistent.

One request supports up to 3 JPEG, PNG, or WebP reference images. References can communicate layout, information hierarchy, or visual direction. Do not upload protected material you cannot use, patient identifiers, or other sensitive data.

Prompt Structure

Prompt pattern

Create a [figure type] about [scientific subject].
Use [layout] with [number of stages or panels].
Include labels: [required labels].
Style: [visual style and background].
Use case: [paper, review, slide, grant, poster, protocol].
Avoid: [unwanted details].

Revision Strategy

After the first draft, make smaller requests. It is usually better to revise one kind of issue at a time.

Goal	Better instruction
Preserve layout	"Keep the same three-panel layout, but make the icons cleaner."
Correct text	"Replace 'T cell activation' with 'T cell priming'."
Reduce clutter	"Remove decorative particles and keep only the pathway components."
Clarify flow	"Make arrows larger and ensure the sequence moves left to right."
Fix one area	"Redraw this receptor as a clean membrane receptor icon."

When to Regenerate

Regenerate when the overall concept, composition, or scientific story is wrong. Revise when the figure is close but needs style, label, clarity, or local corrections. Move to Vectorization when the remaining work is precise cleanup, alignment, element movement, or export.

Review scientific accuracy

AI-generated figures can look polished even when a label, arrow, or relationship is wrong. Always review scientific accuracy before sharing.

Next: Edit text and redraw regions, upscale or create a white background, or vectorize the image.

What This Module Does

Image Generation is where you create the figure. It helps turn a scientific idea into a visual draft, then gives you ways to revise that draft without starting over every time.

Use this module when you want to:

create a figure from a written description;
use a sketch or reference image to guide layout and style;
improve an existing figure draft;
correct labels or terminology;
redraw one confusing area;
make the image clearer before export or vectorization.

Starting Points

Text to Figure

Describe the scientific story, layout, labels, and intended use. This is the fastest way to create a new graphical abstract, mechanism diagram, or workflow.

Sketch to Figure

Use a rough drawing when the structure matters more than visual polish. A sketch can communicate panel order, arrows, and object placement.

Reference to Figure

Add examples, screenshots, or previous drafts when you want to preserve a visual direction, level of detail, or composition.

Figure Refinement

Start from an existing draft and ask FigRay to simplify, clean, relabel, or clarify it.

Best Fit

Figure type	What to provide
Graphical abstract	Study summary, desired stage order, main labels.
Mechanism diagram	Key actors, interactions, arrows, labels, and direction.
Protocol schematic	Step list, order, grouping, and output format.
Slide figure	Main message, audience, and desired level of detail.
Existing draft cleanup	Current image plus what should stay and what should change.

Basic Workflow

Describe the figure goal in one or two sentences.
Specify layout, panels, flow direction, and required labels.
Add sketches or reference images when visual guidance matters.
Generate the first draft.
Review scientific accuracy, composition, and text.
Use follow-up instructions for global revisions.
Use focused edits for labels, local regions, clutter, and clarity.
Export directly, or move to Vectorization when the figure is close to final.

Models and Reference Images

Prompt Structure

Prompt pattern

Create a [figure type] about [scientific subject].
Use [layout] with [number of stages or panels].
Include labels: [required labels].
Style: [visual style and background].
Use case: [paper, review, slide, grant, poster, protocol].
Avoid: [unwanted details].

Revision Strategy

After the first draft, make smaller requests. It is usually better to revise one kind of issue at a time.

Goal	Better instruction
Preserve layout	"Keep the same three-panel layout, but make the icons cleaner."
Correct text	"Replace 'T cell activation' with 'T cell priming'."
Reduce clutter	"Remove decorative particles and keep only the pathway components."
Clarify flow	"Make arrows larger and ensure the sequence moves left to right."
Fix one area	"Redraw this receptor as a clean membrane receptor icon."

When to Regenerate

Review scientific accuracy

AI-generated figures can look polished even when a label, arrow, or relationship is wrong. Always review scientific accuracy before sharing.

Next: Edit text and redraw regions, upscale or create a white background, or vectorize the image.

What This Module Does

Starting Points

Text to Figure

Sketch to Figure

Reference to Figure

Figure Refinement

Best Fit

Basic Workflow

Models and Reference Images

Prompt Structure

Revision Strategy

When to Regenerate

Table of Contents

Image Generation Module

What This Module Does

Starting Points

Text to Figure

Sketch to Figure

Reference to Figure

Figure Refinement

Best Fit

Basic Workflow

Models and Reference Images

Prompt Structure

Revision Strategy

When to Regenerate

Table of Contents