SeaArt Guide
How to SeaArt AI - Office Tutorial
How to SeaArt AI - Office Tutorial
  • ✨How to Use an AI Image Generator
  • ✨1-SeaArt AI Basic Page
  • ✨2-SeaArt AI Basic Function
    • 2-1 Text to Image
    • 2-2 Image to Image
    • 2-3 ControlNet
    • 2-4 AI Apps
      • How to publish as App
      • Swift AI Apps
    • 2-5 AI Characters
      • How to create your own character？
      • Character Description Writing Tips
      • Conversation Tips
    • 2-6 Models
    • 2-7 Post
    • 2-8 AI Video Generation
      • Txt2Vid
      • Img2Vid
      • Camera Movement
      • Start and End Frames
    • 2-9 AI Audio
    • 2-10 Workflow
      • Text to Image Workflow
      • IMG2IMG+Partial Repainting
      • Core Nodes
      • Tips
    • 2-11 Canvas
    • 2-12 LoRA Training
      • Flux Lora Training
  • ✨3-Advanced Guide
    • 3-1 Principles of AI art
    • 3-2 LoRA Training (Advance)
      • How To Create Dataset For Training
    • 3-3 Workflow Guide
      • Image Conversion
      • Inpainting
    • 3-4 Canvas Guide
    • Composite Poster
  • ✨4-Parameters
    • 4-1 Model
    • 4-2 Mode
    • 4-3 Basic Settings
    • 4-4 Advanced Config
    • 4-5 Advanced Repair
    • 4-6 Complete Prompting Guide
    • 4-7 Prompt Edit | Keyword Blending Guide
  • ✨5-Practical Examples
    • AI Influencer
    • LOGO Design
    • E-commerce Poster
    • How To Make Multiple Characters
    • Prompt Templates
    • How to Maintain Character Consistency
    • Useful Prompts
  • ✨6-Permanent Events
    • SeaArt.AI Creator Incentive Program
      • Creator Incentive Program FAQ
    • High-Quality Models Recommendation
      • SeaArt Infinity
      • Stable Diffusion 3.5
      • SeaArt Realism
      • NOOBAI XL
      • T-Ponynai3 V6
      • Counterfeit V3.0
      • Temporal Paradox Mix
    • High-Quality AI Apps Recommendation
    • High-Quality Character Recommendation
    • High-Quality Workflow Recommendation
  • ✨7-FAQ
Powered by GitBook
On this page
  1. 3-Advanced Guide

3-1 Principles of AI art

Learn the principles of AI art creation: from text prompts to image generation using text encoders, U-NET, and VAE.

Previous3-Advanced GuideNext3-2 LoRA Training (Advance)

Last updated 9 months ago

The creative process: Enter prompts - Select model - Adjust relevant parameters - Generate image.

The process principle: Prompts - Text encoder - U-NET - VAE - Normal image.

After entering the prompts, a text encoder processes them into embedding vectors. Simultaneously, a random seed generates a noise image. The parameters set for AI art, such as sampling method, sampling steps, image size, etc., are fed into the noise predictor. The noise predictor removes image noise, gradually revealing the image. Finally, the output is passed through a VAE to produce an image discernible to the naked eye.

✨
How SeaArt AI Image Generation Works