Case study 2027: Generative AI for image creation

"A diffuse vision": the pre-released case study for Paper 1 Section B, for exams in May and November 2027. Visionary Studios, a design company, is choosing between diffusion models, GANs and hybrid models to create advertising images, and has to manage the computing cost and the ethical and legal risks.

1. Read the bookletThe official 8-page case study (PDF). Read it twice. 2. NotesEvery section explained with diagrams, plus the research you need beyond the booklet. 3. Lab: activitiesGenerate images from noise, train a real GAN, explore latent space, curate a dataset. 4. TerminologyAll the booklet's terms in exam-ready words, with flashcards and a quick quiz. 5. Exam preparationHow Section B works, what each challenge needs, practice questions with mark points. 6. Exam practiceSection B style questions from the question bank: answer, then mark yourself with the mark scheme, or sit a timed Section B. Quiz 1Terminology and main concepts (SL or HL). Progress is saved on this computer. Quiz 2A second, timed quiz on the terms and concepts.

SL and HL

	SL	HL
Study	Everything except GANs and hybrid models	Everything: diffusion models, GANs and hybrid models in depth; VAEs and flow-based models broadly
Challenges	2: diffusion's denoising and computing demands; ethics of training data (IP and bias)	4: the SL two, plus balancing GAN generator and discriminator; evaluating hybrid models
Section B	12 marks	24 marks

On the notes, lab, terminology and exam pages, Hide HL in the menu hides the HL-only parts.