Product Image Creator
Turn one product brief into three distinct storefront-cover directions with GPT Image 2: consistent brand colors, readable hierarchy, configurable size and quality, and local PNG output you can inspect.
Version 1.2 is live
The current Agensi buyer download and public listing use GPT Image 2. Version 1.2 includes the 1280×720 storefront default, supported custom sizes, quality controls, actionable API errors, and no-partial-set output described below.
One brief, three compositions
Provide a product name and description, then optionally add a price, brand colors, style, canvas size, and quality. The workflow makes three separate GPT Image 2 requests so the result is a useful choice among directions, not three renamed copies of one image.
- Minimalist: restrained layout, clear type hierarchy, and subtle geometric detail.
- Bold: high-contrast gradient, dynamic shapes, and stronger marketplace-grid impact.
- Screenshot: a product-interface mockup direction using laptop and phone framing.
Storefront-sized output without a hard-coded dead end
The default is three 1280×720 PNGs, matching Gumroad's current recommended cover dimensions. For a square store thumbnail, request 1024x1024. GPT Image 2 also accepts auto and flexible sizes within its published limits, and the script validates those limits before spending an API request.
Output files are named {product_slug}_cover_1.png through _3.png and saved to generated-covers/ unless you choose another directory.
What version 1.2 changed
- Replaced the deprecated DALL-E 3 / GPT Image 1.5 route with GPT Image 2.
- Made the actual selected model, size, and quality visible in progress output.
- Added flexible size validation and
low,medium,high, orautoquality. - Added separate guidance for invalid credentials, missing model access, policy refusal, rate limits, server errors, timeouts, and network failures.
- Removed silent deprecated-model fallback.
- Made saving transactional: an incomplete three-image generation run writes no partial cover set.
Requirements and data boundary
You need Python, an OPENAI_API_KEY for a project with GPT Image 2 access, outbound network access to api.openai.com, and permission to write the chosen local output directory. The workflow does not require browser access and does not upload the saved cover files to a storefront.
The $15 skill purchase and OpenAI API usage are separate charges. Quality and size affect generation cost and latency; start with medium for finished covers or low for cheaper exploratory runs.
What still needs a human review
Generated typography can still contain spelling or layout errors, and a plausible mock interface is not evidence that the depicted feature exists. Review every cover at full size and thumbnail size. Check the product name, price, logo use, claims, contrast, crop, and any trademarked material before publishing.
The workflow creates new images; it is not a layered design-file editor, template marketplace, background remover, or automatic Gumroad/Lemon Squeezy uploader.
Questions
What does Product Image Creator generate?
It generates three distinct PNG cover options from one product brief. Each option uses a different composition while preserving the requested style and brand colors.
Which image model does it use?
Version 1.2 uses GPT Image 2 by default. It does not silently fall back to deprecated image models.
What size are the generated covers?
The default is 1280 by 720 pixels. You can request 1024 by 1024 for square thumbnails, auto sizing, or another size that satisfies GPT Image 2 constraints.
What access does the skill need?
It needs Python, outbound network access to the OpenAI Images API, the OPENAI_API_KEY environment variable, and write access to the chosen local output directory. It does not need browser access.
Does the purchase include OpenAI API usage?
No. The skill purchase and OpenAI API billing are separate. You supply an API key for a project with GPT Image 2 access.
What happens if one of the three requests fails?
The run reports the relevant authentication, access, safety, rate-limit, API, or network problem and leaves no partial cover set behind.
Create a cover set you can compare, not a one-shot prompt.
Version 1.2 is live on Agensi for $15 or 75 credits. Review the displayed permissions and refund terms before buying; OpenAI API usage is separate.
Get Product Image Creator on Agensi