Codex Pet v2 explained: actions, directions, and package structure
A plain-language guide to the 9 action rows, 16 look directions, spritesheet, and pet.json structure used by Codex Pet v2.
DeskPeto delivers a complete resource that ChatGPT/Codex can identify and play—not a loose collection of pictures. Knowing the basic structure makes installation and troubleshooting much easier.
What is inside a companion package
The two core files are pet.json and spritesheet.webp. The configuration describes the companion, version, and animation regions; the spritesheet stores every frame in a fixed grid. They must stay together.
Renaming, independently compressing, or cropping one file can prevent ChatGPT/Codex from reading the animations correctly.
Nine standard animation groups
Pet v2 uses nine animation groups to represent idle, movement, work, waiting for input, completion, and other states. Together they create a companion that responds to work rather than looping a single idle animation.
DeskPeto keeps face, clothing, proportions, and colors consistent across rows and checks that states are visually distinct.
Sixteen look directions
The companion may face content from different positions on the desktop, so the resource covers sixteen look directions. More directions also create more chances for ears, hair, markings, and accessories to drift.
The approved key character acts as the identity reference used to validate every direction.
The 1536×2288 sheet and installation checks
The complete spritesheet uses a 1536×2288 canvas and a fixed grid. Preserve the original dimensions and transparency; do not re-save the file through a messaging app.
If the companion does not appear after refresh, check the folder nesting, filenames, and pet.json before reopening the Pets menu. The DeskPeto installation guide shows the full sequence.