AI Image Generation Course

Lesson 1 • Core Generative Model Types

Surveys GANs, VAEs, and diffusion models at a conceptual level. Helps learners match model types to appropriate use cases.

Lesson 2 • Ethical and Legal Landscape

Introduces copyright, consent, and bias concerns specific to AI imagery. Frames responsible practice as a professional baseline.

Lesson 3 • History of Generative Visual AI

Traces the evolution from early neural art to modern diffusion models. Provides context for understanding current capabilities and limitations.

Lesson 4 • How Models Learn Visual Patterns

Explains training data, loss functions, and iterative learning in plain terms. Connects model behaviour to the data it was trained on.

Lesson 1 • Downloading and Managing Models

Covers model repositories, file formats, and local storage organisation. Prevents common errors caused by mismatched or corrupted model files.

Lesson 2 • Choosing a Generation Platform

Compares web-based interfaces, desktop apps, and code-based pipelines. Guides learners to the platform that fits their workflow.

Lesson 3 • Running Your First Generation

Executes a basic text-to-image generation with default settings. Confirms the environment works and introduces the generation feedback loop.

Lesson 4 • Hardware and Cloud Requirements

Outlines GPU, RAM, and storage needs for local setups and cloud alternatives. Enables informed decisions about infrastructure investment.

Lesson 1 • Iterative Prompt Refinement

Establishes a systematic test-and-revise workflow for prompt improvement. Builds diagnostic skills to identify why outputs deviate from intent.

Lesson 2 • Positive and Negative Prompting

Teaches how to specify desired elements and explicitly exclude unwanted ones. Reduces artifacts and off-target content in outputs.

Lesson 3 • Prompt Weighting and Syntax

Introduces token weighting, attention syntax, and keyword ordering. Gives precise control over which elements dominate the composition.

Lesson 4 • Anatomy of an Effective Prompt

Breaks down subject, style, medium, lighting, and mood as prompt components. Shows how each element steers model output.

Lesson 1 • Resolution and Aspect Ratio

Covers native resolution limits, aspect ratio effects, and tiling strategies. Prevents common distortion artifacts from incorrect resolution settings.

Lesson 2 • Batch Generation and Variation

Uses batch size and seed variation to explore the output space efficiently. Supports rapid concept exploration and selection workflows.

Lesson 3 • Parameter Presets and Profiles

Organises frequently used parameter combinations into reusable presets. Speeds up production workflows and ensures consistency across sessions.

Lesson 4 • Steps, CFG Scale, and Seed

Explains how inference steps, classifier-free guidance, and seeds interact. Provides a mental model for reproducible and varied outputs.

Lesson 1 • Inpainting Workflow

Covers mask creation, fill modes, and prompt alignment for inpainting. Enables targeted edits without regenerating the entire image.

Lesson 2 • Image-to-Image Fundamentals

Introduces denoising strength and how input images guide generation. Establishes the conceptual difference between txt2img and img2img workflows.

Lesson 3 • Outpainting and Canvas Extension

Extends image boundaries using outpainting to expand scenes or compositions. Teaches overlap and prompt strategies for seamless edge blending.

Lesson 4 • Style Transfer via Img2Img

Applies new artistic styles to existing images while preserving structure. Connects img2img parameters to controlled stylistic transformation.

Lesson 1 • Combining Multiple ControlNets

Stacks conditioning inputs to control pose, depth, and edges simultaneously. Teaches weight balancing to prevent conflicting control signals.

Lesson 2 • Depth and Normal Map Conditioning

Extracts depth and normal maps to preserve 3D spatial relationships. Maintains scene geometry when restyling or transforming images.

Lesson 3 • Edge and Line Conditioning

Uses Canny, MLSD, and lineart preprocessors to guide composition from edges. Enables precise layout control from sketches or existing images.

Lesson 4 • ControlNet Architecture Overview

Explains how ControlNet attaches to a base model to inject spatial guidance. Provides the conceptual foundation for all conditioning workflows.

Lesson 1 • Textual Inversion Embeddings

Trains a new token to represent a concept using a small image set. Produces lightweight, portable embeddings for style or subject injection.

Lesson 2 • LoRA Training and Application

Trains low-rank adaptation layers to capture styles or characters efficiently. LoRAs are the most practical fine-tuning format for most workflows.

Lesson 3 • Dataset Preparation and Captioning

Covers image curation, resolution standardisation, and automatic captioning tools. High-quality datasets are the single largest driver of training success.

Lesson 4 • Dreambooth Full Model Training

Runs full Dreambooth fine-tuning for deep subject or style embedding. Covers prior preservation loss to maintain general model capability.

Lesson 1 • Designing a Repeatable Pipeline

Maps the full generation process from brief to final asset using structured stages. Reduces rework by standardising decisions at each pipeline step.

Lesson 2 • Post-Processing Integration

Integrates AI outputs with raster editing tools for final polish and correction. Bridges the gap between generation and delivery-ready assets.

Lesson 3 • Quality Assessment Criteria

Defines objective and subjective quality metrics for AI-generated images. Enables consistent evaluation across team members and projects.

Lesson 4 • Version Control and Asset Management

Organises prompts, seeds, models, and outputs for traceability and reuse. Prevents loss of successful configurations and supports team collaboration.