RepDB
← All posts

Can You Use a Free Exercise Dataset Commercially? A 2026 Licensing Guide

June 29, 2026 · 5 min read · by Sergei Argutin

Most teams grab a free exercise dataset, ship it, and never read the license. Here's an honest, accurate breakdown of which exercise datasets are actually safe for commercial use in 2026 — public-domain options, restricted APIs, image-provenance traps — and when a licensed dataset is worth paying for.

If you’re building a fitness app, the first thing you need is exercise data — names, muscles, equipment, instructions, and ideally images. The second thing you need, and the thing almost nobody checks until a lawyer or an app-store reviewer asks, is: am I actually allowed to ship this?

I went through this myself building RepDB. So here’s the honest version — not “all free data is illegal, buy mine,” but what each common option actually permits, where the real traps are, and when paying for a licensed dataset is the right call.

Not legal advice. Licenses change. Always read the current license/ToS of any dataset before you ship, and when in doubt, ask a lawyer.

The one-minute answer

SourceTypical licenseData — commercial use?The catch
free-exercise-dbUnlicense (public domain)✅ data · ⚠️ imagesThe JSON is clearly Unlicense; the photos trace to an older repo that never explicitly licensed its images — verify before a paid product
wrkout/exercises.jsonPublic domain✅ data · ⚠️ imagesOlder, less structured; same image-provenance question
Kaggle “exercise” scrapesOften unspecified / scraped⚠️ RiskyImage provenance unknown — the data being “free” ≠ the images being cleared
Hosted exercise APIs (free tiers)Proprietary ToS / metered⚠️ DependsRedistribution and offline bundling are usually restricted — read the ToS
RepDB free tierAttribution license✅ Yes (in-app, with attribution)400 exercises, flat 512×512 images only; no redistribution as a dataset, no generative-AI derivation
RepDBExplicit commercial license✅ Yes (paid tier)It costs money — that’s the trade for a clear license + support

Short version: genuinely public-domain exercise data exists and you can ship it commercially — but check the images separately; that’s where the risk hides. The decision isn’t “free vs legal” — it’s whether the free option’s hidden costs (below) are cheaper than a licensed one for your product.

”Free” that’s genuinely free

Some open datasets — like free-exercise-db and the older wrkout/exercises.json — release their code and JSON data into the public domain (Unlicense). For the text layer that’s as clean as it gets: use it commercially. The images deserve one extra look, though: free-exercise-db’s photos trace back to the older wrkout repo, which never explicitly licensed its images — the maintainer can dedicate his own work to the public domain, but whether the upstream photos were his to dedicate is an open question. For a hobby project, fine; for a paid product, check the chain of title.

What you’re also trading away is everything around the license:

For a side project that’s completely fine. For a product you’re charging for, those gaps become your engineering backlog.

”Free” that isn’t quite free

Two categories quietly get teams into trouble:

Commercial APIs with a free tier. Several exercise APIs let you call them for free but their Terms restrict redistributing or bundling the data offline — which is exactly what you do when you ship an app. The data being reachable isn’t the same as it being licensed for your product. Read the ToS for the words “redistribute,” “derivative,” and “cache.”

Scraped datasets (often on Kaggle/GitHub). A dataset can be posted “free” while the images inside it were scraped from a source that never licensed them. The text might be fine; the images can be a copyright claim waiting to happen. If a dataset can’t tell you where its images came from, treat that as a risk — especially for the images, which are the part most likely to be someone else’s IP.

This is the trap that matters most: “the dataset is free” and “the images are cleared for commercial use” are two different statements. Verify the second one.

When a licensed dataset is worth paying for

Pay for a licensed exercise dataset when one or more of these is true:

That’s the gap RepDB fills. It’s a curated, commercial-safe exercise dataset: 455 fully illustrated exercises, two image styles (classic transparent + flat), EN/DE/ES, delivered as JSON + SQLite + WebP — one payment, no API, no rate limits, with an explicit commercial license.

Try before you decide

You don’t have to take my word for it — the data shape and image quality are the whole point, so look first:

If a public-domain set covers your needs, use it with a clear conscience. If it doesn’t, now you know exactly what you’re paying for — and what to check before you ship either way.