ExerciseDB vs free-exercise-db: Hosted API or Open Dataset?

First, which ExerciseDB?#

The exercisedb.dev home page now reads "Previously ExerciseDB" above the name AscendAPI, and AscendAPI's documentation, read on October 22, 2026, describes two products:

• ExerciseDB V1. A free hosted tier of "1,500 exercises with 180p GIFs. No authentication required", at oss.exercisedb.dev/api/v1/exercises, plus a paid RapidAPI tier with 360p to 1080p images and GIFs. Records carry bodyParts, targetMuscles, equipments and a gifUrl.
• ExerciseDB V2. On RapidAPI only, with an MP4 video per exercise and a stated 11,000+ exercises. Its free Basic plan gives 2,000 requests a month and 200 exercises, with watermarked media.

An older ExerciseDB listing on RapidAPI, published by justin-WFnsXH_t6 and covered by the ExerciseDB integration guide, also exists; we could not read it today or establish whether it is the same product. AscendAPI's paid prices and terms of use were out of reach too, and none of it is self-hostable: when we cloned its GitHub repository on October 22, 2026 it held only a README and an AGPL-3.0 LICENSE.

So ExerciseDB is an API you call, free to start, with the full catalogs on paid plans. free-exercise-db is the opposite model: files you download and own.

ExerciseDB vs free-exercise-db at a glance#

For the cost side across the whole category, see exercise database API pricing.

What is actually in free-exercise-db#

We downloaded the combined file the README points to and counted it on October 22, 2026, rather than repeating the README's "800+". Every record has id, name, force, level, mechanic, equipment, primaryMuscles, secondaryMuscles, instructions, category and images, validated against a JSON Schema in the repository.

By level, 525 are beginner, 294 intermediate and 57 expert. The gaps are documented: the README lists force, mechanic and equipment as incomplete in some files, and in our count those fields were null in 30, 87 and 77 records respectively. Five records have an empty instructions list, and the README notes "a small number of duplicate images". Muscles come from a fixed list of 17 values (biceps, lats, quadriceps and so on), and IDs are strings built from the name, such as Alternate_Incline_Dumbbell_Curl.

Serving it takes no backend. The README's documented shortcut is to fetch the combined JSON from raw.githubusercontent.com and prefix each image path with the repository's exercises/ raw URL:

curl -O https://raw.githubusercontent.com/yuhonas/free-exercise-db/main/dist/exercises.json
# an image path such as Air_Bike/0.jpg becomes:
# https://raw.githubusercontent.com/yuhonas/free-exercise-db/main/exercises/Air_Bike/0.jpg

For production, copy the files to your own storage or CDN rather than hot-linking GitHub. The repository also has Makefile tasks to rebuild the combined file and to emit newline-delimited JSON for a PostgreSQL import, and a searchable Vue front end at yuhonas.github.io/free-exercise-db for browsing.

On provenance, the README says the data was restructured from wrkout/exercises.json, and that upstream repository's LICENSE.md is also the Unlicense. Neither README we read names who made the original photographs; if image provenance matters to your legal review, that is the question to ask.

Where ExerciseDB pulls ahead#

Media and scale. If your UX depends on a moving demonstration, ExerciseDB is the only one of the two that serves it: animated GIFs on V1, free at 180p, and on V2 a record with a videoUrl pointing at an .mp4, plus overview, exerciseTips, variations, keywords and relatedExerciseIds that free-exercise-db has no equivalent for. V2 also names muscles anatomically ("Pectoralis Major Clavicular Head" in its sample) and tags a gender for its demonstrations. And it is a live service: someone else hosts it, adds exercises and fixes data, and V1's free tier lets you try that without a key.

Where free-exercise-db pulls ahead#

Ownership and zero obligations. The Unlicense dedicates the work to the public domain, so you can ship the data and images in a closed-source commercial app with no attribution, no share-alike and no copyleft. There is no key to leak, no quota to hit, no per-request bill, and no terms of use that can change under you. Because the whole dataset is about a megabyte of JSON, you can bundle it into an app and search it offline. The cost is that you own everything a service would do: hosting, search, fixes, and filling the null fields.

Which should you pick?#

• You need GIF or video demonstrations: ExerciseDB. V1 for GIFs, free at 180p, or V2 for MP4 video on RapidAPI plans; read the terms that govern the media files before you ship them.
• You need data you can own, with no license strings: free-exercise-db. It is the most permissive exercise source this site has reviewed.
• You are prototyping: either works for free. free-exercise-db works offline and has no terms to read; ExerciseDB V1's free tier needs no key but comes with terms we could not read. Move to a paid plan once you know whether motion is worth paying for.
• You want an open catalog with an API and translations instead: neither; look at wger, with its AGPL code and per-entry Creative Commons data. The wger alternatives page compares it against both of these.

Moving between them#

It is a field-by-field re-map, not a swap. free-exercise-db's primaryMuscles uses 17 broad group names; ExerciseDB V1 uses targetMuscles arrays such as pectorals, and V2 anatomical names; the older RapidAPI listing used a single target in our July review. Equipment labels differ in vocabulary and case (body only and e-z curl bar in free-exercise-db against LEVERAGE MACHINE in the V2 README sample). IDs share nothing, so keep a mapping table keyed by your own IDs. Media is the expensive part: going from ExerciseDB to free-exercise-db means two still images per exercise where you had motion, and going the other way means checking that the exercises your users already logged exist in the new catalog. For other ways out of either product, see ExerciseDB alternatives.

Watch the fine print#

Our counts describe free-exercise-db's main branch as we downloaded it on October 22, 2026; a later commit can change them. ExerciseDB's catalog and media figures are AscendAPI's own, its free tiers are as its documentation described them that day, and its paid prices and terms sit on pages we could not read from our environment, so verify them before you budget.