Generate AI

Generate AI

Create images, chat with LLMs, and generate media using The Grid's decentralized AI network.


Den (電) - Your Credits

Den is the platform currency for AI generation.

  • New users get Den when signing up
  • Workers accrue more Den for contributing GPU power
  • Spend Den to generate images, text, and media

We're currently giving generous amounts of Den so you can explore freely.


Generate Images

aipg.art

aipg.art (opens in a new tab) is the AI art gallery powered by The Grid.

What you can do:

  • Generate images with Flux, SDXL, Stable Diffusion, and more
  • Browse community creations
  • Use pre-built workflows or customize parameters
  • Download your generations

How to use:

  1. Go to aipg.art (opens in a new tab)
  2. Sign up or log in
  3. Enter your prompt
  4. Choose a model and settings
  5. Generate

Available Models:

ModelBest ForVRAM on Worker
FluxHighest quality, photorealism12GB+
SDXLHigh-res, detailed images8GB+
SD 1.5Fast generations, wide LoRA support6GB+

Parameters you can adjust:

  • Steps (quality vs speed)
  • CFG scale (prompt adherence)
  • Sampler and scheduler
  • Dimensions
  • Seed (for reproducibility)

NFTs (Coming Soon)

Mint your generations as on-chain NFTs with full reproducibility.

Unlike typical NFT JPEGs with IPFS links that can rot, Grid NFTs store all generation parameters on the blockchain:

  • Model hash
  • Prompt and negative prompt
  • Seed, steps, CFG, sampler
  • Workflow recipe

Anyone with the model can regenerate the exact image from chain data alone. Future-proof, verifiable, truly ownable AI art.


Chat with AI

aipg.chat

aipg.chat (opens in a new tab) is an OpenWebUI-powered chat interface running on The Grid.

What you get:

  • Chat with open-source LLMs (Llama, Mistral, Qwen, etc.)
  • No corporate API middleman
  • Decentralized inference via Grid workers
  • Conversation history

How to use:

  1. Go to aipg.chat (opens in a new tab)
  2. Sign up or log in
  3. Select a model
  4. Start chatting

The responses come from real GPUs running real models—no API wrappers, no OpenAI under the hood.


Use the API

For programmatic access, use The Grid API directly.

Get Your API Key

  1. Go to dashboard.aipowergrid.io (opens in a new tab)
  2. Create an account
  3. Generate an API key
  4. Check your usage at api.aipowergrid.io (opens in a new tab)

SDK

Use the official SDK for easier integration:

Repository: github.com/AIPowerGrid/grid-sdk (opens in a new tab)

import { GridClient } from '@aipowergrid/sdk';
 
const grid = new GridClient({ apiKey: 'your-api-key' });
 
// Generate an image
const result = await grid.generate({
  prompt: 'a robot painting a sunset',
  model: 'flux-schnell',
  width: 1024,
  height: 1024
});
 
console.log(result.imageUrl);

Discord Bots

The Grid powers several Discord bots you can add to your server:

Image Bot

Generate images directly in Discord.

Repository: github.com/AIPowerGrid/grid-discord-image-bot (opens in a new tab)

News Bot

AI-powered news summaries and updates.

Repository: github.com/AIPowerGrid/grid-discord-news-bot (opens in a new tab)

Aigarth Chatbot

Chat with an AI personality in Discord.

Repository: github.com/AIPowerGrid/aigarth-chatbot (opens in a new tab)


What Models Are Available?

The Grid's ModelVault tracks all registered models. Current categories:

Image Generation

  • Flux.1 (Schnell, Dev)
  • SDXL and variants
  • Stable Diffusion 1.5
  • Inpainting models
  • ControlNet (Canny, Depth, Pose)

Text Generation

  • Llama (various sizes)
  • Mistral
  • Qwen
  • Other Ollama-compatible models

Coming Soon

  • Video generation
  • Audio/music generation

Models are continuously being added. The ModelVault on Base mainnet ensures only verified models are available.


Tips

For better images:

  • Be specific in prompts
  • Use negative prompts to exclude unwanted elements
  • Higher steps = more detail (but slower)
  • Same seed + same settings = same image

For faster generations:

  • Use Flux Schnell for quick iterations
  • Lower step counts for drafts
  • Generate at smaller sizes, upscale after

For chat:

  • Larger models give better responses but may be slower
  • Be clear and specific in your questions
  • The model only knows what it was trained on (no real-time info)

Questions?