GenAI School · learn it here, then run it yourself
Start with our one-tap templates and character effects, then graduate: build your own ComfyUI pipelines, run them on Colab / Modal / Fal, pull models from Hugging Face and Civitai, upload your own templates, and share what you make in the forum and the Library.
Basics — how AI images work
How AI image generation works 5 min
A prompt describes what you want; a diffusion model turns noise into an image that matches it. A negative prompt says what to avoid. Seeds make results repeatable. That is the whole game.
Prompt craft (and where good prompts come from) 10 min
Structure: subject, style, lighting, camera, detail; then a negative prompt for what to avoid. You do not have to invent them — huge open prompt libraries exist to learn from.
Our templates — make something in one tap
Use our templates — one tap 5 min
Do not start from a blank box. Pick a template, fill a couple of fields, and generate. We have image templates, CapCut-style reel/video templates, and character-transformation effects.
Characters, consent & public figures 8 min
Make a reusable character on the platform, or use a fictional one — no proof needed. A real person’s face needs consent. Public figures are fair game for satire and art; do not imply they endorse anything, and never make non-consensual intimate images.
ComfyUI — build your own pipeline
ComfyUI — build your own pipeline 15 min
ComfyUI is a free node graph: load a model, encode prompts, sample, decode, save. Once you can read a graph you can do things templates cannot — img2img, inpainting, ControlNet, character consistency, video.
Video & motion — clips, effects, animation
Free clip effects — shatter glass, explode, zoom 4 min
The zero-setup way to get motion: take any image you made and add an epic effect right in your browser — shatter glass, explode, zoom, glitch — then download a clip. No GPU, no account, no cost.
Real AI video with ComfyUI (AnimateDiff / SVD) 15 min
For true generative motion — a still that actually moves, camera pushes, morphs — you use a video model in ComfyUI: AnimateDiff, Stable Video Diffusion (SVD), LTX-Video or Wan. These need a GPU, so you run them on rented or your own compute (next lesson).
Run the video yourself — free & cheap GPUs 12 min
Video models need a GPU. You do not have to buy one: run the ComfyUI video workflow on a free or cheap hosted GPU. Kaggle gives ~30 free GPU hours a week; Colab is ~$10/mo; Modal has free monthly credits; Fal is pay-per-clip. Same models, your compute.
Run it yourself — Colab, Modal, Fal
Run it yourself — Colab, Modal, Fal 15 min
Graduate off our templates: run the same tools on your own compute. Google Colab is the cheapest start (~$10/mo Pro), Modal gives free monthly credits, Fal is fast pay-per-generation. Same models, your control, no lock-in.
Model & prompt libraries — Hugging Face, Civitai, GitHub
Models & LoRAs — the open libraries 12 min
Checkpoints set the base style; LoRAs add a character or look on top. The big open libraries are Hugging Face and Civitai. CHECK THE LICENSE on each — some are free for anything, some are non-commercial, which matters if you sell or mint.
Build on our open repo — use our code in YOUR AI 8 min
Everything here is open. The Bot repo has pieces you can lift straight into your own AI project: the effect / reel / character template registries, the consent gate, the Graphene chain client, and the knowledge corpus — all zero-dependency ESM modules, easy to drop in. Take what helps, build your own, contribute back.
Share — upload, Library, forum
Upload your own template 8 min
Made something good? Package it as a template — a name, the prompt/workflow, and the fields others fill in — and share it. Browse and remix what others uploaded, gallery-style.
Publish to the Library & the forum 6 min
Write up what you made — the template, how it works, the prompt — as an article in the Library of Ashurbanipal, then start a forum thread that links to it. That is how the good stuff gets found and built on.
NFTs & collectibles — the honest version
NFTs & collectibles in game economies 12 min
You can mint a template or image as an NFT — useful as an in-game item, a collectible, or a way to sign your work on-chain. First, read this honestly.
Honest heads-up: most templates or images you mint will have little or no market value — treat them as numismatic or sentimental keepsakes, not investments. A few may become rare collectibles over time; none are guaranteed to, and most will be worth less than what you put in. We teach how NFTs work in game economies and collections — this is education, not financial advice or a promise of value. Never spend money on minting that you cannot afford to treat as a hobby.