Submissions

[tender portal]

Submissions here!

[alpine dagger]

yoooo

[left echo]

Spent money so you can maybe win the money back 💀

[ornate lintel]

yeah how are they not providing credits to play with it

[ripe venture]

fr

[junior echo]

fr

[brittle lichen]

Hi, OpenAI. I would be interested, however it seems clear to me that API access is not being made available freely for the purposes of this competition, therefore limiting what I am willing to test and as such while I could write up some code I would not be enabled to test it as I am unwilling and unable to spend money to develop a program for use with your API to use as part of a contest to get money. It would be a gamble of both time and money and that is not something I personally would be willing to do. It’s a shame, too, because when I saw this I got excited to build some of the ideas I’ve had for this kind of tool but alas I do not think OpenAI will be the company who’s API I end up making use of for these projects at the end of the day due to this.

[tender portal]

Should move convos to the #dev-chat so this is free for submissions. Happy to respond more there.

[glacial island]

Submissions

[latent iris]

image-gen-1.5 or gpt-image-1.5?

[grim maple]

No free tokens = no free work

[pseudo stratus]

give $20 participation api credits

[brittle lichen]

Actually everyone I’ve discussed and found something better another user mentioned

From @earnest axle : if in your API account settings you enable "share data for training" you will recive up to 10million model tokens(per day) (model dependant), perhaps that miught help?

[bleak meadow]

OpenAI worth $500B, gives $75 for first prize in a competition you have to pay to enter 😂

[cinder leaf]

the data sharing says this, so I don't think it applied to gpt image 1.5. I do notice that it doesn't mention gpt-5.2 here, yet I recieve it through the API for free. I'll guess i'll do a test with 1.5 image

1.5 image is NOT part of the data sharing program. I just blew my life savings of 17 cents to tell you this

[novel gate]

Some heroes don't wear capes.

[vernal lance]

I switched to img gen it’s so fantastic I love it so much I would choose that over my will to live
https://github.com/notvhalan/frontendbibliotec/settings

[rapid pelican]

elo

[vital blaze]

😓
I would like to enter but… if it’s not free. Might as well dont do it all.

[brittle lichen]

I want an allowance of $500 a day, a stipend(idk what that is but I heard it in a movie once) of like $1000/week and I want a room and board budget of $10,000/month and I want also a company credit card. For this, I will make one (1) program that literally just allows the user to enter a prompt and receive an image result back. /jk

[inner sorrel]

https://github.com/strato-space/media-gen-mcp

Media Gen MCP — Strict TypeScript MCP server for OpenAI gpt-image-1 and sora-2: generate, edit, and fetch images/video with smart compression, resource_link and image outputs, and a built-in test tool for MCP clients. Production-focused (full strict, zero any, lint), optional sharp processing.

---

Media Gen MCP is a strict TypeScript Model Context Protocol (MCP) server for OpenAI Images (gpt-image-1) and OpenAI Videos (Sora): generate/edit images, create/remix video jobs, and fetch media from URLs or disk with smart resource_link vs inline image outputs and optional sharp processing. Production-focused (full strict typecheck, ESLint + Vitest CI). Works with fast-agent, Claude Desktop, ChatGPT, Cursor, VS Code, Windsurf, and any MCP-compatible client.

Design principle: spec-first, type-safe image tooling – strict OpenAI Images API + MCP compliance with fully static TypeScript types and flexible result placements/response formats for different clients.

Generate images from text prompts using OpenAI's gpt-image-1 model (with DALL·E support planned in future versions).
Edit images (inpainting, outpainting, compositing) from 1 up to 16 images at once, with advanced prompt control.
Generate videos via OpenAI Videos (sora-2, sora-2-pro) with job create/remix/list/retrieve/delete and asset downloads.
Fetch & compress images from HTTP(S) URLs or local file paths with smart size/quality optimization.
Debug MCP output shapes with a test-images tool that mirrors production result placement (content, structuredContent, toplevel).
Integrates with: fast-agent, Windsurf, Claude Desktop, Cursor, VS Code, and any MCP-compatible client.

[weak drum]

haha 1st price 75$ is such a joke

[inner sorrel]

but honor >> price.

[inner sorrel]

This is a Community Challenge, not an OpenAI sponsorship.

p.s. new release media-gen-mcp coming soon.

[brittle lichen]

Also you’re the first I’ve see actually submit anything

Idk about you but I value people who actually contribute and what they say over others such as myself who have not done so

[echo kiln]

I just fixed the image with gpt-image 1.5 😆

[brittle lichen]

I know I’ve said this enough and people still won’t get it but:

1. it’s a community event not a official OpenAI one

2. they’re doing a show off your code as open source type deal, not a buyout or even use of your code. I assume you can just license it non commercial and bam you’re still good and they can’t use your work if you’re so worried about that.

3. imagine demanding more money for a community event tbh, I could never. Sure, I addressed concerns about ability to participate due to needing to spend on ai use for this particular model, but that was also addressed and came to terms I find unacceptable for myself and many might even, but I’m not gonna make demands of change. I just hope the next challenge uses a model that is accessible via the free credits from allowing training use of inputs and outputs.

[brittle lichen]

Can you fix it to say GPT-Image 1.5 API Challenge?

[inner sorrel]

https://media-gen.stratospace.fun/media/output_1766008739_media-gen__openai-images-edit_9cc76754-8d4f-4dd5-897e-bc51dd466366.png

7500 cent challenge )))

[runic rain]

Hybrid workflow using Image Gen as a high-fidelity canonical asset origin.
https://github.com/Janksuu/canonical-image-origin

[brittle lichen]

Beautiful stuff

[raw aurora]

Are we able to change the prize? Or just creds

[brittle lichen]

thank you both

[tender bridge]

$75 not worth my time!!

[runic rain]

But replying in here to tell us that was worth your time?
Just a note, but you don't need to spend money to simulate an api with Python. If anything, you could get a relatively accurate token count and cost estimate.

[pseudo stratus]

lol

[latent iris]

@tender portal You still have "Image Gen" mentioned 3 or more places in this post.

[cyan cave]

My tool already supports image generation and can be used with the new model. You can install with uv tool install zapgpt and then use in command line like : zapgpt -p openai -m gpt-image-1.5 -ip "Create image of a hacker" -is 1024x1024 . Github repo is at: https://github.com/raj77in/zapgpt

[shrewd oar]

so you have to pay to join the challenge essentially since the API (especially image gen) isnt free and you have to test the project??

[midnight hawk]

you also have to pay for your computer, your electrictity, etc so yes...

[latent iris]

Here is an app me and Chat GPT 5.2 made called Sketch2Model. I'll worry about the github stuff later.

[tender bridge]

This is cheap way for big tech to steal people ideas!

[maiden heath]

https://tenor.com/kT6IdeuxBE3.gif

[brittle lichen]

License your work under a noncommercial license then?

[inner sorrel]

https://github.com/strato-space/media-gen-mcp

Media Gen MCP is a strict TypeScript Model Context Protocol (MCP) server for OpenAI Images (gpt-image-1.5, gpt-image-1), OpenAI Videos (Sora), and Google GenAI Videos (Veo): generate/edit images, create/remix video jobs, and fetch media from URLs or disk with smart resource_link vs inline image outputs and optional sharp processing. Production-focused (full strict typecheck, ESLint + Vitest CI). Works with fast-agent, Claude Desktop, ChatGPT, Cursor, VS Code, Windsurf, and any MCP-compatible client.

v. 1.1.0

[sleek dagger]

Hey everyone, here’s my Dev Challenge submission: Note Polisher (gpt-images-1.5)

It turns messy notes into a clean, one page study sheet you can actually review. Upload a screenshot or paste text, pick 1 of 5 styles, and it generates a polished, print ready output image.

Repo: https://github.com/Yazan10x/note-polish
Demo: https://youtu.be/U4gpXMrDsnE

How I use Image Gen 1.5: it transforms the note into a structured study sheet image with clear hierarchy (and optional diagrams depending on the preset picked).

Accessibility: Large text + spacing, bold text, high contrast, and reduced motion (all saved in LocalStorage).

Run locally:
bash ./setup_local.sh sk-your-openai-key
bash ./run_local.sh

[covert cosmos]

ty

[blazing pollen]

This should be the winner lol

[glossy jacinth]

https://github.com/Cosmic-Game-studios/Brandkit-Forge

[echo kiln]

👍

[soft gyro]

Good project. I wish I could mix AI, random code, and my own knowledge in the way you just did there. It still looks "Frankeinstenish", but so does every hackathon project.
Hackathons/chalenges/events are great for learning. I remember the last hackathon I attended in University of Florida. It was my first time making a Chrome extension, and we won! Third place, but still, not bad for my first Chrome extension. Lol.
As long as you enjoy the journey and learn, go for it! But don't be mislead: this is a corporation that wants your repo for its own good, not yours, you don't owe them anything... Unless you work for CloseAI - Great for you if you do! But see what Ilya Sutskever said about this corporation. This project is not Linux, this is already in Microsoft's hands.
So, yeah, that is the most objective way I can describe this situation. And I can go on ranting about Nihilism, career and human development, etc, etc. But the last thought I want to imprint is: I admire the love and effort that you have put into your project.
PS: I believe some of your dependencies require a more updated version of NodeJS, so, make sure the engines in package.json matches your local node -v.
glhf

[shut valley]

https://github.com/aqua2777/StickerCraft

[rough marsh]

LocaleLens — AI-powered localization for marketing visuals - made just for this contest!

https://github.com/dsj7419/localelens

I built a tool that transforms marketing screenshots into localized variants using a three-model pipeline: GPT-4o Vision analyzes the image and detects text, GPT-4o translates and writes image-specific prompts, and gpt-image-1.5 generates the localized output with streaming preview.

My stepson is a professional translator fluent in Spanish, French, and Arabic. I wanted to show him how AI is approaching the localization space he works in every day. LocaleLens supports all three of those languages, including full RTL Arabic rendering.

Key features:

• Works with any image (not just pre-configured demos)
• Pixel-perfect composite mode achieving 0% drift outside masked regions
• Real-time streaming preview during generation
• Translation verification using GPT-4o Vision re-read

the README has a lot of gif's and png's you can see as well. I did put some recommendations and findings at the end of the README as well.

Built with Next.js 15, tRPC, and runs entirely locally with just an OpenAI API key.

[oblique blaze]

https://github.com/MoonLayOfficial/VideoGenStudio

VideoGen Studio turns a topic into a structured plan, illustrated chapters, and narrated audio, then compiles everything into a smooth final video. It emphasizes step-by-step control: users review and edit the plan before generating assets. The pipeline uses gpt-5.2 for planning with web search, gpt-image-1.5 for visuals, and gpt-4o-mini-tts for narration.

[loud scroll]

https://github.com/kimaher/AI-Branding-Proj

[wraith scaffold]

Hey everyone!
Over the last couple of days I put together a small prototype — MysticArtifact.
A tiny destiny path where you can summon your own unique artifact (You will receive two versions the artifact, stylized and realistic).
Every artifact comes out different, shaped by the path you choose and wrapped in its own little story.

I’d love to see what your fate weaves.
If you want, share your results here — I’m genuinely curious what kinds of artifacts will appear for others.
I’ll attach a few of my own below!

P.S.
Under the hood it’s all powered by gpt-5-nano + GPT Image API…
but shhh… you didn’t hear that from me. ✨🔥😉

P.P.S.
MysticArtifact was forged through a three-step “model alchemy”:

1. gpt-5.1-instant → the ignition spark that formed the first concept.
2. gpt-5.2-pro → the deep crucible where the prototype took shape.
3. gpt-5.2-codex → the refining flame that completed the architecture and polished the final build.

Happy new year 2026!🎄
https://github.com/1dZb1/MysticArtifact

[ember mirage]

https://github.com/a55hot/DungeonGPT

DungeonGPT is a compact AI companion for tabletop RPGs: it quickly creates characters and story hooks, runs sessions in chat, and enhances them with illustrations and audio.

[alpine dagger]

?

[tame glacier]

https://github.com/Terraforming-Planet/Graphic-gen-Terrain-Formation-planet-Photovoltaic-Vehicles
🌍 Terraforming-Planet | Image-Gen API – Live Terraforming & Photovoltaic Vehicles

Hello everyone 👋
I’m submitting my project to the Image-Gen API Challenge.

🔵 Repository
Click repository here:
here

🟠 Issues / Development & Deployment
Click issues here:
here

🟢 Live Frontend (GitHub Pages)
Click live demo here:
here

🟣 Live Image-Gen API (Cloudflare Worker)
Click API here:
here

🚀 About the project
Terraforming-Planet is an experimental project using the OpenAI Image Generation API (gpt-image-1.5).
The goal is to generate realistic and educational visualizations focused on:
planetary terraforming such as mountains, valleys, and terrain formation,
photovoltaic-powered machines and vehicles,
self-sustaining energy concepts,
and visual learning for environmental and planetary engineering.

🧠 Goal
The main goal is to combine AI image generation, education, and experimentation,
showing how modern APIs can support environmental thinking, sustainable technology,
and future planetary engineering concepts.

🛠️ Technical status
The project runs locally and is also deployed on a live production server.
A public frontend is connected to a working image generation API.

🧩 Tech stack
OpenAI Image Generation API,
Cloudflare Workers,
HTML, CSS, and JavaScript.

The project is actively developed and open to further experimentation.
Feedback is very welcome.

Good luck to all participants.

[silver gull]

https://github.com/KennethJAllen/proper-pixel-art

Convert noisy, high resolution pixel-art style images from Image Gen to clean, true-resolution pixel-art assets.

Repo here

Image Gen API instructions here

[oblique blaze]

nice

[shut valley]

Nice repo

Quick question though is the "terraforming" part more conceptual?
From the code it looks like a general image generator

[tame glacier]

Yes — for now the terraforming part is mainly conceptual and visual. The code is a general image generator on purpose, used as a visual sandbox for terrain shaping, water flow, microclimate, and renewable-powered machinery. There is still time until the end of the challenge, so updates and additional generator features may be added.

[tame glacier]

Thank you, I make bonus
B64 generation (Mode B) is back: the generator returns valid base64 images again.
Archive/KV behavior is safer: when the KV binding is missing or misconfigured, the app doesn’t crash — it disables saving and shows a clear message.
Added Mode D (Scene Editor): basic scene controls (select object, move with arrows, light-angle slider, simple touch/drag rotation).
Added a Mirror object option to place a second reference object next to the selected one.
Added Texture from URL (with a CORS warning) + apply/clear controls.
Added a “3D Model Generator” tab: quick presets (pyramid / car / excavator / tree / mountain) that generate prompts and attempt to replace the current “cube” conceptually.
Added AI Replace Selected (polish + save): takes the current scene view and sends it through image editing, then optionally saves the result to the archive.
UI includes a credit line: Built for the OpenAI Community Dev Challenges.
But: it still needs more work — not everything behaves like a real 3D editor yet (selection, touch rotation, mirror placement, archive reliability, and “true 3D object generation” are still not as solid/accurate as expected).

[tender portal]

Hey @vernal lance this github repo seems private so cannot properly review it with the team. LMK if you can open it today? If not, understandable too.

[tender portal]

🏆 Community Dev Challenge: Image Gen API — Winners

Huge thanks to everyone who submitted projects!

🥇 1st Place — BrandKit-Forge @glossy jacinth
Stood out for its clean execution and strongest overall use of the Image API.

🥈 2nd Place — LocaleLens @rough marsh
An ambitious entry that made solid use of both image generation and edits.

🥉 3rd Place — Proper Pixel Art @silver gull
A fun and creative supporting tool that the judges really enjoyed.

Congrats to the winners 🎉

Winners will receive their prizes, and as a bonus, everyone who participated in our first challenge will receive a small amount of API credits.

Expect a DM from @turbid wedge soon! Do not give information to others attempting to message you for details.