Each row in this table represents a media item (e.g., an image or a sound) attached to a sprite in the most recent version of a project shared on the Scratch website at the point of data collection. Users could import their own media into Scratch from their hard drive, or even from their webcam or microphones, using the Scratch authoring environment. A single project could have many pieces of media. Because the software used to extract this data failed to parse some projects, it is not available for every project included in the projects table.
This table includes text input directly from users. Due to technological error, data corruption, and intentional action by users, this table includes invalid UTF-8 data. For completeness sake, we have included these data. Using the find_fix_invalid_encoding_in_json.py program included in the example scripts to decode the JSON data, we found that there were 0 rows with text encoding issues.
This table includes 44,205,193 observations. Each row of this table corresponds to a single piece of project media.
| Variable | Type | Description |
|---|---|---|
| project_id | integer | Unique identifier for the project (see the projects table). |
| name_base64 | character | Visible name of the piece of media given by Scratch automatically or by the user manually. Unfortunately, this texst was input with unknown text encoding. Rather that try to deal with data with unknown encodings, we have opted to encode the binary data reflected in the database using base64 encoding. |
| size | integer | Size of the media element in bytes. |
| media_type | factor | Type of media, it can be either image or sound. |
| from_library | logical | Boolean variable that is true when the media element comes from the Scratch library. |
A density plot showing the distribution of media items by size is as follows:

The distribution of media items by from_library is as follows:
| variable | TRUE | FALSE | NAs |
|---|---|---|---|
| from.library | 4064742 | 40140451 | 0 |
In terms of media_type, 36,785,568 (83%) media elements are images and the remainder are sounds.
We omit 15,999,895 observations because they refer to observations in older, non-public, versions of projects that were replaced on the site. Additionally, we omit 12,229,841 observations from projects which are not included in our projects table (e.g., they were removed from the site by their creators).
These data are access restricted. Details on gaining access are provided in the "Permission to Use Datasets" section of the overview page.