Approve the look first
Attach a figurine photo and describe the colors and details to keep. Review the cartoon character before making the full animation set.
Not quite right? Adjust the character first.The introduction and documentation are available in five languages. This does not mean five-language CLI output or guaranteed generation quality in every language.
OPEN SOURCE SKILL 0.1.0 preview
ToyPal · Your Desktop Companion
Your favorite figurine,
now a desktop companion.
Start with one photo and keep the features you love. Approve the character, create its animations, then install the companion you choose in Codex.
Free, open-source code · Requires Codex image capability, hatch-pet and generation quota
01 / Start with a photo
At every step,
you decide.
Approve the character before generating all animations. This may reduce a complete redo, but success on the first try and fixed completion times are not guaranteed.
See how it moves
Check edges and animation continuity, then preview the movements. Passing technical checks still leaves visual approval up to you.
Animation problems? Repair them; do not skip checks.Install your companion
When the preview looks right, say “Install this pet”, then select it in Codex's pet picker. Existing folders with the same name are preserved, not overwritten.
Installed does not mean selected.START IN CODEX
Install the dependencies and skill, attach a photo, then enter:
Use $toypal to turn this figurine into my desktop pet. Preserve its main colors and decorations; a cartoon redesign is welcome. Let me approve the character before generating animation, then give me a preview and installation package.
02 / Is it for you?
Like the idea?
Check what you need.
A local tool for people who already have Codex image capability, not a standalone Mac or Windows desktop-pet app.
One clear figurine
Use a clear, full-body photo of one opaque figurine, preferably front-facing or at a three-quarter angle. The result is a cartoon redesign, not a guaranteed pixel-exact cutout.
A working generation setup
You need Codex image generation capability and quota, official hatch-pet, Python 3.10+ and Pillow. Running the scripts alone cannot generate a character.
Time to review and adjust
Reference images may go to your image service. Generation and retries consume quota. Approve the character first and stop if you are unhappy, rather than retrying indefinitely.
Can I use it without Codex?
Not currently. There is no hosted photo generator or independent pet runtime. Read the installation guide and check your setup before starting; no registration or payment is needed on this page.
What if generation, animation or installation fails?
Correct the character baseline first if the look is wrong. Return to generation and repair for motion or transparent-edge issues. Preserve existing folders when installation names conflict. Read the check, preview, package and install instructions ↗
03 / Current evidence
One existing example,
not a promise for every figurine.
An existing local bird package was used to verify v2 atlas checks, 11 animation GIF previews, ZIP export and installation into an isolated folder. Installed files were checked for equality.
These validate the package tools, not a complete photo-to-pet generation run with a new figurine. Previews use fixed 150ms frames, not actual Codex playback timing.
- Complete generation with a new figurine: unverified
- Windows on real hardware and independent user experience: unverified
- No universal success rate, fixed completion time or zero generation cost promised
04 / Photos and licensing
Your photos
are not public by default.
Original photos are not included in exported pet ZIPs or published by default. Reference images may still be sent to the image service you use. Share photos and characters only if you have the rights to do so.
The workflow and package tools use MIT; some validation tools retain Apache-2.0. Independently maintained, not an official OpenAI or museum project.
START WITH A FIGURINE YOU LOVE
Give your figurine
a new way to keep you company.
Tell us the failed step, retry count, installation result and why you kept or closed it after seven days. You do not need to publish the original photo.