My "Image (sequence) Player" in the WLED-MM-P4 "Anniversary Edition" for the ESP32-P4 which I'm currently putting the finishing touches on.
Any USB disk inserted into the P4 is loaded on mount and up to 80% of PSRAM (generally 32MB) is used for image sequence caching. JPEG data is cached directly to PSRAM and then decompressed with P4's on-chip accelerated JPEG decoder on the fly. The Pixel Processing Accelerator is used for accelerated playback with real-time scaling and other effects.
This is also demoing the P4's Parallel IO output to drive 16 pins at once. There's some glitching during the image caching, but once things are cached it's instantaneous to switch patterns and it's also glitch-free. Here it's 16x512 pixels per chain, and runs much faster than 30 FPS as it's limited to here. 128x64 running at up to ~97 FPS with WS281x overclocking.
The glitches only seem to be related to these framebuffer effects - regular WLED effects run fine without glitches while the caching happens in the background.