What is this?
GPT Image Playground is a web UI for generating and editing images through the OpenAI gpt-image-2.5 API, with support for OpenAI-compatible endpoints, sub2api, fal.ai, and custom HTTP providers.
GPT Image Playground is a web UI for generating and editing images through the OpenAI gpt-image-2.5 API, with support for OpenAI-compatible endpoints, sub2api, fal.ai, and custom HTTP providers.Is this a ready-to-use web app or a library I plug into my own project?It is a standalone web app built with React/Vite/Tailwind that you deploy (or run locally) and use directly in the browser, not a library to embed in code.
It is a standalone web app built with React/Vite/Tailwind that you deploy (or run locally) and use directly in the browser, not a library to embed in code.How do I get it running?Deploy it with one click on Vercel, publish it to GitHub Pages via the included GitHub Actions workflow, or deploy static assets to Cloudflare Workers with Wrangler; a Docker setup is also provided.
Deploy it with one click on Vercel, publish it to GitHub Pages via the included GitHub Actions workflow, or deploy static assets to Cloudflare Workers with Wrangler; a Docker setup is also provided.Do I need an OpenAI API key to use it?You need an API key from OpenAI or a compatible provider (or fal.ai/sub2api) to generate images; the app itself only requires Node/npm if you build it yourself.
You need an API key from OpenAI or a compatible provider (or fal.ai/sub2api) to generate images; the app itself only requires Node/npm if you build it yourself.Can I point it at a local or intranet API without HTTPS?The Vercel demo's bound domain enforces HTTPS for API calls due to security policy, so non-HTTPS local/intranet APIs won't work there — use the GitHub Pages build or self-host instead.
The Vercel demo's bound domain enforces HTTPS for API calls due to security policy, so non-HTTPS local/intranet APIs won't work there — use the GitHub Pages build or self-host instead.What does Agent mode do?Agent mode is a multi-turn conversational generation flow built on the Responses API: it keeps context, can reference images from earlier turns with @, batch-generates multiple related images in one turn, and lets you branch by editing or regenerating a message.
Agent mode is a multi-turn conversational generation flow built on the Responses API: it keeps context, can reference images from earlier turns with @, batch-generates multiple related images in one turn, and lets you branch by editing or regenerating a message.How does transparent background generation work?In gallery mode with PNG/WebP output you can enable transparency either natively (asking the model to return an alpha channel, if supported) or via local post-processing, where the model renders on a solid green/magenta backdrop that the browser then removes.
In gallery mode with PNG/WebP output you can enable transparency either natively (asking the model to return an alpha channel, if supported) or via local post-processing, where the model renders on a solid green/magenta backdrop that the browser then removes.How many reference images can I use when editing?You can upload up to 16 reference images via clipboard or drag-and-drop, plus use a built-in mask editor for targeted edits.
You can upload up to 16 reference images via clipboard or drag-and-drop, plus use a built-in mask editor for targeted edits.Can I connect providers other than OpenAI?Yes — besides OpenAI-compatible endpoints, it supports sub2api (async), fal.ai (queue-based), and any custom HTTP provider you import via a JSON configuration.
Yes — besides OpenAI-compatible endpoints, it supports sub2api (async), fal.ai (queue-based), and any custom HTTP provider you import via a JSON configuration.What are similar projects?storycast, wardrobe, youtubepro, infinite-canvas, scira, printfilm, crm, RedInk
storycast, wardrobe, youtubepro, infinite-canvas, scira, printfilm, crm, RedInkWhich topics is it in?gpt-image, vercel, image-editing, cloudflare-workers, text-to-image, chat, tailwindcss, vite, image-generation, generative-ai, local-first, cloud
gpt-image, vercel, image-editing, cloudflare-workers, text-to-image, chat, tailwindcss, vite, image-generation, generative-ai, local-first, cloud