How to Use GPT-Image 1.5: A Step-by-Step Guide

Key Takeaways

• GPT-Image 1.5 works best when you treat it like a structured design assistant, not a “magic art button,” especially for clear marketing visuals, diagrams, layouts, and posters that need precision.

• Choose text-only mode when you’re generating from scratch, and switch to image+text when you need consistency, controlled edits, or you want to preserve identity and layout while changing just one element.

• Lock your aspect ratio early based on the channel you’re designing for, and use higher quality settings whenever you need crisp typography, dense layouts, or small labels that must stay readable.

• Use a prompt that the model can follow: define the output type, describe the subject and scene, add composition and style cues, then state constraints like “no extra text” or “change only X” to reduce randomness.

• Iterate like a pro by changing one variable at a time, and for edits, repeat your “keep everything else the same” invariants (framing, lighting, subject identity) on every pass to prevent drift.

1. Prompt: Photorealistic product shot

This prompt delivers a premium product photo with realistic textures and studio polish.

Prompt Template:

Prompt: 
Goal: Photorealistic product photo for ecommerce and ads.
Subject: [PRODUCT] on a bed [SURFACE].
Material/finish: [MATERIAL], with realistic texture and subtle imperfections.
Lighting: soft studio lighting with a gentle rim light, realistic shadows and reflections, clean background.
Composition: centered subject, shallow depth of field, minimal props, premium look.
Constraints: no text, no watermark, no logos, no extra objects.

2. Infographic/diagram

Educators and YouTube Automation Entrepreneurs thrive on scannable visuals for tutorials or faceless thumbnails. This creates a hierarchy-driven infographic with perfect legibility that is ideal for course graphics or process breakdowns.

Prompt Template:

Prompt: Create a vertical 2:3 infographic titled "[TITLE]" with 5 numbered steps.
Each step includes: a simple icon on the left, a bold heading, and 1–2 short lines of supporting text.
Style: clean, modern, high readability, white background, subtle accent color [ACCENT COLOR], consistent spacing.
Constraints: all text must be perfectly legible, no extra text, no watermark, no logos, no clutter.

3. Text in image

GPT Image 1.5 excels here with verbatim text rendering such that promo assets come ready with flawless branding, and no warped headlines or typos.

Prompt Template:

Prompt: Design a minimalist poster (2:3) with a clean background and strong typography.

Headline text (EXACT, verbatim, no extra characters): "[HEADLINE]"
Subhead text (EXACT, verbatim, no extra characters): "[SUBHEAD]"

Typography: modern sans-serif, high contrast, clean kerning, centered alignment.
Layout: headline top third, subhead below it, generous whitespace.
Constraints: text must appear once, perfectly legible, no watermark, no logos.

4. “Change only X” edit prompt

Precision edits save hours for consistent branding by swapping backgrounds or recoloring without drift.

Prompt Template:

Prompt: Edit the provided image.

Change ONLY: [THE ONE THING YOU WANT CHANGED].
Keep everything else the same, including: subject identity, facial features (if present), lighting direction, shadows, camera angle, framing, background layout, and overall image quality.
Do not add any text, logos, or watermarks. Do not introduce new objects.

5. Prompt: Multi-image compositing

Composite product shots or scene builds effortlessly grab an object from one image, drop into another with matched lighting. Composite images are a great way to layer multiple themes into one finished product.

Prompt Template:

Prompt: Use Image 1 as the base scene.
Take the [OBJECT] from Image 2 and place it into Image 1 at [LOCATION IN FRAME].
Match lighting direction, perspective, scale, and contact shadows so it looks naturally photographed.
Do not change anything else in Image 1. No text, no watermark, no logos.

A. Step 1: Log in and choose your model

Start by logging in to your invideo account and clicking on the ‘Agents & models’ tab.

Once there, select the ‘GPT Image 1.5’ model and create a new project. If you are creating something from scratch, use text-only. This is ideal for quick concepting, product scenes, posters, thumbnails, and simple marketing visuals. If you want control and consistency, however, use image+text (edit/multimodal).

B. Step 2: Pick aspect ratio + quality

Keep the aspect ratio decision tied to where the image will live: Use 1:1 for feed posts and general assets, 2:3 for vertical layouts like infographics and poster-style designs, and 3:2 for wide, presentation-like visuals.

C. Write a prompt the model can follow

The fastest way to get “predictable” results is to structure your prompt.

This is how you move from “pretty” to “usable.”

D. Generate → iterate with one change at a time

Generate your first version, then do one of two things:

If you’re close, adjust one variable at a time: “make the lighting warmer,” “increase spacing,” “reduce background clutter,” or “use a more minimal font.”

If you’re editing an image, repeat the constraints every time. A pattern that works well is: “Change only X. Keep everything else the same,” followed by a short list of what must not change. That “preserve list” is the difference between consistent iterations and frustrating drift.

Here is an example of a prompt and its corresponding outcome on GPT Image 1.5 on invideo:

Prompt: 
Goal: Photorealistic product photo for ecommerce and ads.
Subject: Luxury watch on a bed of pebble shaped rocks.
Material/finish: Rose Gold, with realistic texture and subtle imperfections.
Lighting: soft studio lighting with a gentle rim light, realistic shadows and reflections, clean background.
Composition: centered subject, shallow depth of field, minimal props, premium look.
Constraints: no text, no watermark, no logos, no extra objects.

Outcome:

Also Read: 80+ AI Image Prompts for Effortless Visual Creation