Add materials for "How to Generate Images With the OpenAI API"

[realpython-bot]

Sample code for the upcoming tutorial How to Generate Images With the OpenAI API (CMS post 2285).

This replaces the workflow in openai-dalle/, which no longer runs: OpenAI retired dall-e-2 and dall-e-3 on May 12, 2026 and removed the image variations endpoint at the same time. Calls to those models now return a 400, and /v1/images/variations returns a 404.

Contents

File	Purpose
create_png.py	Create an image from a text prompt and write it straight to a PNG
create.py	Create an image and save the whole JSON response
convert.py	Decode a Base64 JSON response into PNG files
edit.py	Edit a saved image with a follow-up prompt

Notes for reviewers

• All four scripts were run against the live API with gpt-image-2, not just adapted on paper.
• create.py, convert.py, and edit.py are byte-identical to the corresponding code blocks in the tutorial.
• convert.py carries over from openai-dalle/ unchanged — the response shape for saved JSON is the same.
• edit.py replaces vary.py. The variations endpoint is gone, so editing with a prompt is the closest equivalent.
• The GPT Image models always return Base64 data. There's no response_format parameter and no URL option, so the decode step is now required rather than optional.
• images.edit() needs a (filename, bytes) tuple. Passing raw decoded bytes the way vary.py did fails with an unsupported-mimetype error, since the API infers the file type from the filename.

openai-dalle/ is left in place — say the word if you'd like it removed or marked deprecated in a follow-up.

🤖 Generated with Claude Code