Hi OpenAI team,
I’ve been experimenting with a custom ChatGPT pet named Rook, and it gave me an idea for expanding the pet feature in a way that could make it much more personal, expressive, and fun to use.
The core idea is to let users assign custom pet animations to the different activities ChatGPT is already performing.
Right now, ChatGPT can display activity descriptions such as thinking, working, reviewing, inspecting files, organizing files, inspecting archives/ZIP files, and similar actions. Instead of having many of those activities use the same generic pet animation, users could optionally map a different animation to each activity.
For example, my pet Rook is a small mechanical raven. While ChatGPT is working, he might write in a notebook. While inspecting files, he could use a magnifying glass. While organizing the Library, he could drag boxes around or sweep papers off the desk. While inspecting a ZIP file, he could examine a little box with a giant zipper on it. A failed task could make him faceplant on the keyboard, while a successful task could trigger a short victory animation.
The important part is that the pet would visually react to what ChatGPT is actually doing, making it feel less like a static mascot and more like a tiny digital companion working alongside you.
I also think the creation process could be made accessible to users who know nothing about sprite sheets or animation. A pet editor could simply show something like:
Idle: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Thinking: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Working: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Reviewing: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Organizing: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Inspecting: Frame 1 | Frame 2 | Frame 3 | Frame 4 | Frame 5
Users who want complete control could upload their own frames. Users who do not could upload one reference image of their character and ask ChatGPT to generate the animation automatically—for example, “This is my pet. Make him organize files by carrying boxes.”
A preview button, animation-speed control, loop/play-once option, and fallback animation would make the system straightforward without requiring users to understand the technical implementation.
There could also be a natural difference between basic and expanded pet customization. A basic pet could have a small set of states such as Idle, Thinking, Working, and Sleeping. Paid plans could optionally unlock the larger activity-animation library, additional custom reactions, more animation slots, and AI-assisted animation generation.
That way the free/basic experience would still feel complete, while users who really enjoy customization could make their pet feel much more alive.
I think this could help pets become something people become genuinely attached to rather than another cookie-cutter assistant mascot. Two users could perform the same task in ChatGPT while having completely different characters and reactions on screen. That personality—and the funny moments people would inevitably share—could become part of what makes the feature special.
Thanks for considering it. I’ve had a surprising amount of fun building Rook, and seeing the activity states change while ChatGPT works made me realize how much potential there is if users could customize those reactions directly.
— C. L. Atkins