General Settings

TABLE OF CONTENTS

• Iterations
• Steps
• CFG Scale
• Model
• Scheduler
• VAE
• Seed
• Ratio
• Denoising Strength 

Iterations

• The number of iterations of the prompt that will be ran. This means for prompt utilizing dynamic prompts, the number of images produced will be (# of dynamic prompts) x (# of iterations). Iterations started with the same prompt will be added to the queue as part of the same batch.
• Tip: Iterations can be utilized to easily generate multiple options with a given prompt and chose the best one to send to the Unified Canvas for Inpainting / Outpainting

Steps

• Number of steps that will be performed in each generation. Higher step counts will typically create better images but will require more generation time.

CFG Scale

• Classifier Free Guidance (CFG) Scale controls how closely the model will follow your prompt.
• A value of 1 will give the model complete control, while a higher value will be more restrictive. Higher values can sometimes lead to distorted results.

Model

• The Model used for the denoising steps
• There are three base models that are used: SD1.5, SD2.1 and SDXL, with SD1.5 and SDXL being the most popular base models. Many new models have been created using these, with different models specialized in producing certain aesthetics and content. Each base model type is separated in the Model dropdown.
• Adding custom models to InvokeAI:
  • Indie tier users can add custom models to use with their account
  • Enterprise users can contact their administrator to add custom models

Scheduler

• Scheduler defines how to iteratively add noise to an image or how to update a sample based on a model's output
• There are many different Schedulers, each producing different results
• Two popular ones to start with are Euler and DPM++ 2M Karras. Modifying the step count will affect the image result based on the scheduler used. With these two schedulers, Euler can generally be used with a lower step count, while DPM++ 2M Karras will require a higher step count.

VAE

• Variational AutoEncoder (VAE) is the model responsible for translating the model output into an image.
• Different VAEs will produce slightly different outputs with saturation and minor details - For most purposes, the Model’s default VAE will be optimized for that model.

Seed

• Controls the starting noise used for generation. Disable “Random Seed” to produce identical results with the same generation settings.

Ratio

• The ratio of the generated image. Models are best outputting certain image sizes, with each model having a specific ratio:
  • SD1.5 Models perform best at 512x512 images
  • SDXL Models perform best 1024x1024 images
• Generating outside of these image sizes can result in images that have duplications or are distorted

Denoising Strength 

• Denoising Strength determines how much noise is added to the initial image that is passed to the model. A denoising strength of 0 means that the image will remain exactly the same and a denoising strength of 1 will result in a completely new image.