AI
Guide

Generate images locally with InvokeAI

Learn how to locally generate AI images of anything you can imagine with InvokeAI. It's totally free.

Shento Hendriks

By Shento Hendriks

•

Interface of InvokeAI
Interface of InvokeAI

Introduction

InvokeAI is a powerful, open-source application that lets you generate AI images directly on your computer—no internet connection or subscription required. Unlike cloud-based services, InvokeAI gives you complete control over your creative process with features like:

  • Unified Canvas: Create and edit images in one seamless workspace
  • Advanced Prompting: Fine-tune your results with detailed text descriptions
  • Inpainting/Outpainting: Edit specific parts of images or expand them beyond their borders
  • Model Management: Easy installation and switching between different AI models
  • One-Click Installation: Simpler setup than most competitors

Requirements

  • A modern Graphics Card (GPU): Something like an NVIDIA RTX 3050 model or newer is recommended.
  • At least 8 GB of RAM (Memory): More is often better, but 8GB is the minimum.
  • Storage: At least 10GB free space for the application and models

Install

Go to the official page and download the installer for your operating system (Windows, Mac, or Linux). Run the installer and follow the on-screen instructions

info icon

info

The installation process is quite straightforward. Keep the default settings—they're optimized for most users. Choose location for the folder, you can easily access.

Downloading models

1
Navigate to Model Manager

After installation, you'll need to download AI models before generating images. Open InvokeAI and click the Model Manager icon (the square icon, 4th from the left).

2
Installing a Starter Model

Navigate to Starter Models, and you'll see a list of different models. Choose models based on your hardware:

  • If you have a powerful computer: install Flux or SDXL models.
  • if you have a weaker PC: install SD1.X models.
  • Recommended starter: Juggernaut XL v9 (excellent for photorealistic images)

Click on the + icon to install them.

3
Prune clean-up
Do icon
After the installation is done, click on Prune to clear the queue. This is good practice as it will free up resources and clear cache.
info icon

info

Visit civitai.com to browse thousands of community-created models with different artistic styles.

warning icon

warning

civitai.com contains adult content—use filters if needed.

Generating images

Selecting your model

Navigate to the canvas (1st icon), go to Generation > Model, and make sure you selected your downloaded model.

Important settings

The settings of InvokeAI are straightforward once you understand them.

Positive Prompt: Descriptive text that tells the AI what you want to create, including subjects, style, composition, and visual elements.

Example: "A serene mountain lake at sunset, oil painting style, warm golden lighting, reflections on water"

Negative Prompt: Descriptive text specifying unwanted elements, styles, or quality issues to exclude from the generated image.

Example: "blurry, low quality, extra limbs, distorted faces, watermark, text"

Aspect: The proportional relationship between width and height of the generated image, expressed as a ratio.

Seed: A numerical value that determines the randomness pattern for image generation, allowing reproducible results when kept constant.

Prompting

This is what I entered in my Postive Prompt, and then pressed Invoke, which generates the image.

A beautiful fantasy castle during the sunset
with an irish landscape.

The result

After a few seconds, the result was quite interesting. The image generated really beautiful.

Generated Image of Irish Castle
Generated Image of Irish Castle

Advanced Settings

There are also advanced settings, the most impactful one are those under Generation. If you reveal those options you see: Scheduler, Steps, and CFG Scale.

Scheduler

Controls the algorithm that guides how the AI transitions from random noise to your final image. Think of it as the "pace and style" of the generation process.

  • Euler a (Best for beginners due to its speed and efficiency. Works well at lower step counts (15-30 steps). Tends to generate diverse results quickly but may sacrifice fine details. Ideal for brainstorming or quick iterations.)

  • PM Family (DPM++, DPM2, etc.) Widely recommended for high-quality results once you know your prompt. Typically requires 20-40 steps for optimal output. DPM++ 2M Karras is a community favorite for its balance of quality and efficiency. Avoid at low step counts (<15).

  • LMS/DDIM Best for sharp images with well-defined subjects (e.g., portraits). Requires moderate steps (25-35). More consistent than Euler a but slower. Ideal for final renders.

Steps

Determines how many refinement cycles the AI performs during image creation.

  • Higher step counts (e.g., 30-50) allow more iterations to refine details and textures, generally improving quality but requiring more processing time.
  • Lower steps (e.g., 15-25) speed up generation but may compromise detail. The ideal setting varies based on your hardware and image complexity.

CFG Scale

Classifier-Free Guidance Scale determines how strictly the AI adheres to your text prompt.

  • Lower values (1-5): grant the model creative freedom, potentially deviating from your description for more artistic interpretations.
  • Higher values (7-15): enforce tighter prompt alignment but risk producing stiff or unnatural-looking images. Medium values (5-8) typically balance creativity and fidelity.