heatdrop.ai

A MIDI token format that gives language models a sense of musical time, so they can generate structured pop piano instead of aimless note soup.

A Markov chain and an RNN walk into a bar, and the punchline rhymes with "Herve Leger."

Prism is what happens when a wallpaper app refuses to stop adding features.

A bridge between Python ML workflows and openFrameworks for real-time interactive art.

BlendGAN skips the usual per-style retraining by learning to blend face and style representations inside a single GAN.

Goopt uses GPT-3 to generate fake web pages in real time, letting you "browse" a web that doesn't exist.

A local automation layer for Stable Diffusion, VQGAN+CLIP, and Disco Diffusion that turns a text file into hundreds of images while you sleep.

A NIPS workshop paper that taught CNNs to draw pictures with text characters—and left the repo half-finished.

Google Creative Lab's archived experiment lets you duet with a Magenta-trained model that continues whatever you play.

VOCA turns arbitrary speech audio into animated facial geometry for any static character mesh.

SolidUI wants you to describe a graph in one sentence and have it appear—whether it actually delivers is harder to verify than the README suggests.

Most projects fight hallucinations; this one harvests them for raw idea generation.

A Chinese full-stack hobby project bundling video piracy tools, ChatGPT wrappers, and enough feature creep to fill a mall kiosk.

A curated index of 100+ papers on reversing generative models so you can edit images in their native latent language.

An AUTOMATIC1111 extension that turns outpainting into hypnotic infinite-zoom videos without leaving the WebUI.

A templating language that brings conditionals, recursion, and 70+ shortcodes to your AI image prompts.

This extension wires Anthony's QR Toolkit into AUTOMATIC1111's Stable Diffusion WebUI so you don't have to export and re-import images by hand.

A transpiler and runtime that turn ComfyUI's graph workflows into plain Python you can diff, loop over, and run in Jupyter.

A research GAN that reconstructs realistic images from sparse edge and color hints, with an interactive drawing tool attached.

A Jupyter notebook that downloads YouTube videos, asks GPT-4 to find the "viral" bits in the transcript, then crops around faces for vertical shorts.