Roughly in the order they tend to decide the outcome. For each one: what it means, the question to put to a vendor, and our own answer.
1.Catalogue size, and how much of it you will actually show
Headline exercise counts are the easiest number to compare and the least useful one. Gender variants, equipment substitutions and camera angles all get counted differently, so a catalogue three times the size can hold much the same number of distinct movements. Ask instead how many exercises your product will ever surface, and whether it covers the equipment your users own.
How we answer it: 1,943 exercises across 18 equipment categories and 45 primary muscle groups, and every one of them has video. Fewer rows than the largest catalogue among the vendors compared here, with more media in each row.
2.Media type, coverage and quality
This is where exercise APIs differ most, and where the marketing is vaguest. A GIF is not a video. One clip is not several angles. A watermark on the free tier means the free tier cannot ship. So check the format, the number of renditions, whether every exercise has media or only the popular ones, and whether it is watermarked on the plan you actually intend to buy.
How we answer it: 7,766 MP4 clips: 4 for every exercise, male and female, front and side. Every clip carries a matching thumbnail. Plus 15,144 bodymap images (Professional plan or higher). MuscleWiki branding is embedded in every clip on every plan, so the files never change when your plan does, and the URLs never rotate.
3.Language coverage: advertised versus complete
A language list and language coverage are two different things, and it is the list that gets published. Thirty languages where a given exercise exists in four is a worse outcome for your users than a short list with no gaps. Ask for per-language completeness rather than the list, and ask what comes back for a language nobody has filled in yet: an English fallback and an empty field are very different bugs.
How we answer it: 14 languages at parity: every exercise carries its name in all 14 and returns the same instruction steps in all 14, so no language returns a thinner record than English (non-English needs the Growth plan or higher).
4.Licensing and commercial use
Three separate questions hide behind the word "free". Can you use the data in a product you charge for? Can you store it, or must every screen hit the API? And what happens to the copies you already hold once you stop paying? Among the vendors compared here every restrictive pattern turns up: commercial use limited to paid plans, caching permitted only above a certain tier, caching allowed only per device, and media watermarked until you upgrade. Read that before you read the price, because a plan that forbids caching can cost far more in calls than its headline suggests.
Sources: API Ninjas: Exercises API documentation, API Ninjas: pricing page, ExerciseAPI.dev: catalogue and video-demo counts, plans, caching rule, AscendAPI docs: ExerciseDB V2 overview, WorkoutX: terms of service, watermarking, caching and redistributionHow we answer it: Every paid plan includes commercial-use rights from the first call. Two things are required in return: a verbatim disclosure in your Terms of Service or Privacy Policy, and leaving the MuscleWiki branding in the video you display. An in-app credit is encouraged but optional, and never affects your access. Caching is capped identically on every plan: text metadata up to 30 days, thumbnails 24 hours on-device only, video and bodymaps served no-store.
5.Quota model, and what happens at the ceiling
Metering comes in two shapes, and which one you get matters more than the number attached to it. A monthly call quota is a budget: you plan against it and you know the bill. A rate limit is a throughput cap, which shapes your architecture but says nothing about your monthly total. What almost nobody publishes is what happens at the ceiling: a hard 429, a silent slowdown, or an overage charge. Ask, and get it in writing.
How we answer it: 500 a month free (Playground only), then 1,000 at $10 up to 300,000 at $199.99 a month. The quota is a ceiling rather than a meter: no overage billing, 429 with Retry-After once you reach it, and usage alerts at 80%, 90% and 100% of the monthly quota, 80% and 100% on a Custom plan.
6.Authentication, and how you buy
Buying through a marketplace such as RapidAPI and buying direct are different operational decisions, not just different checkouts. A marketplace gives you one invoice across vendors and a proxy in front of the API. Direct gives you a support relationship, terms you can negotiate and one less hop. If your procurement is already consolidated somewhere, that usually settles it.
How we answer it: Direct. Create an account, generate a key in your dashboard, send it as X-API-Key. Keys are yours to rotate, and changing plan never changes the key. For native apps, your server exchanges the key for a short-lived media token and the device plays /stream/ URLs with no header, so the key never ships in the binary.
7.Operational burden: who runs it at 3am
Self-hosting an open-source project costs nothing in licence fees, and it is not free. Someone patches it, someone runs the database and the media storage, and someone is on call at night. A managed API turns that into a line item and hands the risk to a vendor you have to trust. If you go managed, look at your exit as well: the terms covering what you may keep after you stop paying are the ones that decide how hard leaving is.
How we answer it: We run it: no infrastructure, no upgrades, no media storage and no pager on your side. We do not ship a self-hosted build, so operating the software yourself is not something we can offer. Your exit is governed by our caching terms, which state exactly what you may keep when you stop paying.