AI Pets meet Muse
Pablo has found another home: a little round screen running Muse. His face is familiar. His mouth follows the sound. And the animation pack is the same one we already made.
Explore AI Pets for Muse
After bringing Luna into Claude, we wanted to try another place for our characters to live. The Muse ESP32 SDK already has a device interface, audio support and a connection to Muse. Could we give it an AI Pets face?
Yes. AI Pets for Muse replaces the avatar renderer with our existing animation player. Muse keeps its interface and device controls. The pet arrives as an .aipetframes package, complete with its expressions, speaking poses and reactions.
One character, another place to live
The repository includes Pablo, Luna, Moss, Angel, Cyberpunk Shaman and Lavender Dragon. All six are reviewed 240 × 240 animation packs, imported unchanged. Pick one before building the firmware; there is no character picker on the device yet.
You can also make your own with the AI Pets Sprite Skill for Codex. Describe a character, review its face and animations, then bring the finished pack to Muse. Our Codex starter guide walks through that creative part.
The adapter accepts compatible 120 × 120 and 240 × 240 core packs, up to 3,000,000 bytes. Preparation checks the package before building. Some newer pack features are outside this version’s support, so the repository documents the exact compatibility boundary.
The mouth follows the audio
A speaking face needs to know when sound is happening, how strong it is, and when it stops. We feed the level of the decoded playback audio into the pet’s existing mouth stages. Louder and quieter passages choose different openings; silence brings the mouth back to closed. This is audio-level animation, rather than matching individual phonemes.
We checked every input level from 0 to 100 for all six characters through the real renderer. On the physical 1.85B, Muse’s built-in speech sample produced three visibly different mouth shapes. After playback, the mouth returned to exactly the same closed pixels captured before it. A speaker-to-microphone tone check also passed on both microphone channels.
That local speech test matters, but it is only part of a talking companion. Muse currently returns text replies in this integration. Live spoken replies still need a text-to-speech connection, and account pairing and a complete voice conversation remain unverified. The sample demonstrates working playback-driven mouths without claiming that remaining setup is finished.
A little more room for Pablo
On the Waveshare ESP32-S3-Touch-LCD-1.85B, Pablo sits in the center of the 360 × 360 display. We made him larger and kept that size while speaking. Spoken captions have a dark band below his mouth, so a reply does not squeeze the character into a corner.
The repository also builds for the Waveshare ESP32-S3-Touch-AMOLED-1.75C. The 1.85B is the board we flashed and bench-tested for this release. Physical touch gestures, the BOOT button and LCD appearance still need the final hands-on check.
More places for the same little face
The useful part of a portable pet is being able to keep the character as its surroundings change. You can make a face with Codex, refine its small habits, and bring the resulting package into a compatible player. Muse is another place for that work to come alive.
AI Pets for Muse is an independent integration, not a Meta-endorsed device. Muse’s SDK token terms govern service access separately from source licensing, including personal, non-commercial use and restrictions on distributing devices. No service credentials are included.
Explore the repository, make your own character, or meet Luna in Claude.