cd /news/ai-tools/how-i-made-exercise-icons-with-codex… · home › topics › ai-tools › article
[ARTICLE · art-145207] src=alexop.dev ↗ pub= topic=ai-tools verified=true sentiment=↑ positive

How I made exercise icons with Codex and ChatGPT

A developer built a personal workout tracking app over a weekend using OpenAI's Codex for implementation and ChatGPT for image generation, producing 46 exercise illustrations covering 51 of the app's 82 exercises. The icons were generated in a consistent style — black equipment, silver details, three-quarter view, transparent background — after the developer used a screenshot of gym equipment as a reference image, then requested batches of 20 images at a time and previewed them in the Histoire component playground. The developer said the approach replaced drawing the artwork personally, hiring an illustrator, or accepting the style and gaps of an existing icon library.

by read2 min views1 publishedOct 5, 2026
How I made exercise icons with Codex and ChatGPT
Image: Alexop (auto-discovered)

I’ve always wanted my own workout tracking app. Something I can use at the gym and change whenever I don’t like how it works. Last weekend I started building it with AI, using Codex to do the implementation while I decide what I want and review the result.

I think this could become a small series about the things I learn along the way. The first one is about exercise icons, because that part surprised me.

A workout tracker needs exercises. And once you have a long list of exercises, you want pictures to help you recognize them. I wanted to see the bench or machine next to the name.

I started with SVG icons. They worked, but I didn’t like the look. I wanted something closer to the equipment I see at the gym.

So I took a screenshot of one of the exercises from my gym and asked ChatGPT what it would look like as an icon. I really liked the result. It gave me a style I wanted to use for the rest of the app.

I gave that image to Codex and asked it to use the image generation tool for bench press, squat, deadlift, and a few other exercises. I also asked for a preview in Histoire, the component playground I use for the app, so I could check the images before adding them.

The images use black equipment, silver details, a three-quarter view, and a transparent background. Keeping the reference and those details the same helped them look like they belonged together.

Once I liked the first batch, I asked for 20 more. Then another 20. After those batches, I had 46 illustrations covering 51 exercises. Some exercises share the same equipment image. The catalog has 82 exercises, so there was still more artwork to add.

What I like about this is how much work I would have had to do before. I would have drawn the images myself, paid someone to make a matching set, or searched for an existing library and accepted its style and gaps. For a personal workout app, that would have been a lot of time or money.

Now I can find a look I like, use it as a reference, and ask for more images as I add exercises. I still have to review the results, but creating my own artwork feels like a reasonable part of a weekend project.

I’m already thinking about trying short generated videos to explain the movements. I haven’t done that yet. These icons show equipment, not exercise technique. But being able to even consider that for my own little workout app is crazy.

── more in #ai-tools 4 stories · sorted by recency
── more on @openai 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
→ Live at https://your-agent.zahid.host ✓
Get free account → Pricing
from €0/mo · no card required
LIVE [news/how-i-made-exercise-…] indexed:0 read:2min 2026-10-05 · —