Build fitness apps with a dataset of 1,324 exercises and multilingual descriptions
Imagine a screen where searching for an exercise name displays a GIF of the movement alongside step-by-step instructions. This repository is a dataset that consolidates metadata and visual assets for 1,324 exercises. It allows fitness app developers and researchers to use exercise data immediately without collecting it themselves.
Exercise data structure
Each exercise record contains a unique ID, name, category, equipment, and primary and secondary muscle information. Descriptions are provided in ten languages: English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, and French, and are divided into paragraph form and step-by-step array form by language. Every exercise is linked to a thumbnail image and an animated GIF at 180×180 resolution for visual verification. The data is stored in JSON format, and a JSON Schema file is included for validation.
Browser-based verification tools
Two HTML files are included that can be opened directly in a browser without server installation. index.html is an interactive browser that allows real-time searching of the 1,324 exercises and filtering by category, equipment, and muscle group. setup.html is a developer guide that assists with database setup and API integration, allowing you to generate CREATE TABLE statements and 1,324 INSERT statements in the browser for SQL Server, PostgreSQL, MySQL, and SQLite. It also provides API call examples in various languages such as JavaScript, Python, C#, Java, PHP, Go, and cURL, and guides you on how to generate REST API code based on Express.js, FastAPI, ASP.NET Core, Spring Boot, Laravel, and Gin using LLM prompts.
Data distribution and usage scope
Exercises are divided into various body parts, including arms (292), upper body (227), back (203), waist (169), and chest (163). By equipment, bodyweight exercises are the most numerous with 325, followed by dumbbells (294), cables (157), and barbells (154). Approximately 25% of all exercises can be performed without separate equipment, making them suitable for home workout apps. This dataset can be used for fitness app development, machine learning projects for exercise recognition or recommendation, health and wellness research, and educational demos.
License and media usage terms
The code, tools, dataset structure, and description text are released under the MIT license. However, exercise images and GIFs are copyrighted by Gym visual, redistributed at 180×180 resolution, and must follow the conditions specified in NOTICE.md and LICENSE. To reuse the media, you must review Gym visual's terms and obtain a separate license, while maintaining the attribution © Gym visual — https://gymvisual.com/. This repository does not claim ownership of basic exercise content or media.
