What art generators?

Art generators use machine learning to analyze existing datasets and create new pieces of art.

Comparison

BlueWillow Midjourney Stable Diffusion
Type Online service Online service Middleware
Interface Chat (Discord) Chat (Discord) Web/GUI
Online ✅ ✅ âŦ‡ī¸
At home ❌ ❌ âŦ‡ī¸
Free ✅ Very limited âŦ‡ī¸
Paid ❌ (planned) From $10/month âŦ‡ī¸
Images belong to you ✅ ℹī¸ Paid only ℹī¸ ✅
Quality Good Excellent Requires effort
Text prompt ✅ ✅ ✅
Image prompt ✅ ✅ ✅
Extra tools ❌ Limited âŦ‡ī¸

Notable mentions:

  • DALL-E. First public tool. Free with limited capabilities.

Permissions:

  • BlueWillow
    • Others including BlueWillow may reuse images created in public channels. License unclear.
    • Paid tier. Images created in private channels belong to you. No other re-use permitted.
  • Midjourney
    • Free tier. Images owned by Midjourney, granted BY-NC-4.0 license. Others may also re-use.
    • Paid tier. Images created in stealth mode belong to you. No other re-use permitted.
    • Employees of large companies must use paid tier.
  • Stable Diffusion
    • See hosts for terms of service.
    • Software source available. Limited re-use (covers most scenarios).

Stable Diffusion hosts

Stable Diffusion is an open source art generator that anyone can host at home or on the web.

Self-hosted Google Colab Lucid Creations
Type Self-hosted Online service Client/server
Interface Web (local) Python notebook (web) Desktop GUI
Host Requires Nvidia GPU
8GB+ VRAM		 Google Colab Stable Horde public cluster
At home ✅ ❌ Optional
Free N/A ✅ ✅
Paid N/A ? From ÂŖ1/month ℹī¸
Specialist models ℹī¸ ✅ Paid only ✅ (some with kudos only)
In-painting ℹī¸ ✅ ? With kudos
Out-painting ℹī¸ ✅ ? With kudos
Posing ℹī¸ ✅ Paid only With kudos

Notable mentions:

  • Web GUIs for Stable Horde. Typically fewer features than Lucid Creations:

Permissions:

  • Stable Horde
    • No privacy and permission controls on a public cluster.
    • You can earn "kudos" through ratings, sharing GPU and other community tasks.

Details:

  • Specialist models: Include genres, styles, persistent faces and more. These can be merged with base models. Technologies include DreamBooth, LoRA.
  • In-painting: Fix a detail in the image, such as face or hands.
  • Out-painting: Expand the image beyond its borders.
  • Posing: Provide a simple pose or layout. Technologies include ControlNet, depth2img.