Guide: ChatGPT Image 1.5 Starter Pack

Understanding ChatGPT Image 1.5 for Marketing

ChatGPT Image 1.5 represents a significant upgrade for production workflows, particularly when creating and editing marketing assets at scale. The model excels at following specific instructions without creative drift, maintaining brand consistency across iterations, and preserving critical elements like logos and text.

Key improvements over Image 1 include better instruction adherence (less need to "over-prompt"), stronger editing capabilities that preserve visual identity, improved text rendering for infographics and layouts, 4× faster generation speed, and 20% lower costs via API.

A note on tool selection: For ecommerce use cases specifically, Gemini 3 Image (Nano Banana Pro) remains my preferred image generator and editor. However, if you're committed to the ChatGPT ecosystem, Image 1.5 is a significant upgrade and a capable tool for product marketing workflows.

This guide provides systematic approaches for generating and editing product marketing images with professional results.

Core Prompting Structure

Effective prompts follow a consistent specification format rather than loose creative descriptions. Structure your prompts as: scene → subject → key details → constraints, and explicitly state the intended use.

For new images:

• Specify use case (Instagram ad, product page hero, Amazon listing)

• Define scene environment

• Detail subject and main focus

• Describe composition and framing

• Specify lighting direction and quality

• State brand constraints including colors and style

• Add hard constraints (no watermarks, no extra text)

For edits:

• List what changes (only the specific element)

• List what must be preserved exactly (identity, geometry, layout, text, background, angles)

• Specify what to match (lighting, shadows, perspective)

• State what not to add

For text rendering:

• Quote exact copy or use ALL CAPS

• Specify typography and placement

• Spell uncommon brand names letter-by-letter

For multiple reference images:

• Label each input ("Image 1: product packshot... Image 2: style reference...")

• Explicitly state what transfers where

Essential Techniques

Technique 1: The Preserve List (Preventing Drift)

Add a dedicated preservation block to every edit prompt. This is the most effective way to maintain consistency across iterations.

Preserve List Template:

PRESERVE EXACTLY (do not change):
- Product geometry
- Label text + kerning
- Background color [specify hex code]
- Camera angle
- Crop/aspect ratio

This technique aligns with the model's strengths and significantly reduces unwanted variations.

Technique 2: Single-Change Constraints

Combine "Change only X" with "Add nothing else" for maximum control over edits.

Example Structure:

Change ONLY the cap color to matte black.
Add nothing else. No new objects, no text, no watermarks.

This pattern appears consistently in successful workflows for object removal, color changes, and compositing.

Technique 3: Verbatim Text Rendering

For marketing creative with specific copy, force exact rendering using this pattern:

Billboard text (EXACT, verbatim, no extra characters):
"Fresh and clean"

Quote the exact text, specify placement and typography, and explicitly state "verbatim, no extra characters."

Technique 4: Photorealistic Camera Language

Camera and composition terms produce better realism than generic quality descriptors.

Effective terms:

• Eye-level, 50mm focal length

• Soft window light from left

• Shallow depth of field

• Photorealistic, candid style

Avoid:

• Ultra HD 8K

• Hyper realistic

• Generic quality terms without specific technical details

Technique 5: Variant Factory Method

Instead of one complex prompt attempting everything at once, build systematically:

1. Generate clean base image

2. Create variants through single-change iterations

3. Make one modification per turn: "warmer lighting" or "holiday props" or "more negative space left"

This prevents overwhelming the model and produces more reliable results.

Quality and Speed Settings

For iteration and testing:

• Use standard quality

• Lower input fidelity acceptable for speed

For final production:

• Use high quality (especially for infographics and dense text)

• Use high input fidelity for identity-preserving edits (faces, logos, brand elements)

Production-Ready Prompt Templates

Template 1: Amazon Hero Image (Background Removal)

Extract the product from the input image.
Output: transparent background (RGBA PNG), crisp silhouette, no halos/fringing.
Preserve product geometry and label legibility exactly (no re-rendering of label text).
Optional: subtle realistic contact shadow that respects the alpha.
Do not restyle; only remove background and lightly polish.

Template 2: Lifestyle Scene from Packshot

Place this exact product on a modern bathroom countertop scene.
Preserve the product's geometry, label text, and colors exactly.
Camera: eye-level, 50mm look, shallow depth of field.
Lighting: soft morning window light from the left, realistic shadows under product.
Composition: product centered, clean negative space on the right for overlay text (but do not add any text).
Add nothing else that implies a different brand (no other logos).

Template 3: Color Variant Generator

Change ONLY the product cap color to: matte black.
Preserve everything else exactly: bottle shape, label text, label placement, lighting, reflections, background, camera angle.
No additional objects, no new text, no watermark.

Repeat this prompt structure for each colorway needed.

Template 4: Social Ad Creative with Headline

Create a realistic social ad creative featuring this product on a simple gradient studio backdrop.
Headline text (EXACT, verbatim, no extra characters): "SKIN FEELS CLEAN IN 10 SECONDS"
Subtext (EXACT): "Gentle. Fragrance-free. Dermatologist tested."
Typography: bold modern sans-serif for headline, lighter sans-serif for subtext; high contrast; clean kerning; aligned left.
Layout: product on right third; text on left two-thirds; generous margins.
Preserve product label exactly. No watermarks.

For the Complete Framework

This guide covers core prompting patterns and primary production templates. But there is sooo much more.

Access the full SOP here

• PR outreach data says agency domains win opens with sharper subject lines, but brand domains win replies on credibility, and the sweet spot is agency written pitches sent from the brand domain.

• Build a Client Notebook in NotebookLM with sales notes, site copy, transcripts and sitemap, then prompt it for goals, avatars and differentiators so SEO, copy and sales all work from the same source of truth.

• Grüns hit a $500M valuation fast by running six avatars, each with its own landing page and proof stack, so ads feel made for one problem and CAC drops while relevance scales.

• Research says match tone to intent, go consistently positive for entertainment audiences, go balanced for info seekers, and use trust signals to keep credibility high even when you stay upbeat.

• Mediocre content now gets buried, this playbook uses Screaming Frog plus LLM prompts to score E E A T at scale, spot weak pages fast, and turn audits into a repeatable content upgrade system.

• 2026 may become the year of “100% human” marketing as AI slop gets harder to spot, consumer trust cracks, and brands experiment with no AI promises to feel real again.

• AI transformation is becoming theater, the argument is CEOs must use the tools themselves like Shopify’s Tobi, because acquisitions and CAIO titles cannot replace conviction or culture change.

• A simple style cloning workflow, drop an image into Gemini, extract its style as JSON, save it as a reusable context file, and you can recreate the same look on demand.

• OpenAI says ChatGPT has no ads live, but admits promo like suggestions felt too ad like, turned them off, and is working on better controls after subscriber backlash.

• ChatGPT now lets you tune specific personality traits like warmth, enthusiasm, and emoji use, so the assistant can match your vibe without rewriting your whole setup.

• How has the community been using Gemini 3 and Nano Banana pro, tips and tricks.

• YouTube closes 2025 with updates like voice replies expanding, Shorts experiments, AI suggested clips, comment summaries, and Nano Banana arriving in posts, with more creator tooling rolling into 2026.

• OpenAI is rolling out “Your Year with ChatGPT,” a privacy forward annual recap with stats, awards, and generated media, available to eligible users in select markets including the UK.

🎨 Creatify: Generate high-quality marketing videos from a simple product link or text description.

🐙 AI camp - Access multiple LLMs, assistants, and tools, for teams of all sizes.

📧 SaneBox - Read the important emails in your inbox.

🤖 Gobble bot - Digest your website, PDF, YouTube Video, Notion pages and more into one text file to upload to your GPT chatbot

🦔 Keepi - A personal knowledge assistant

💡 Visual Electric allows users to be inspired by its library of stunning imagery and prompts, and remix their ideas through iteration.

😍It was great, all I needed

🙂Good, not great

🙄It sucked