AWS Certified AI Practitioner AIF-C01 Practice Question

Diffusion models are specifically designed for image synthesis. They iteratively refine random noise into coherent pictures that closely match a provided text prompt, producing high-quality, high-resolution outputs. Transformer-based large language models excel at generating and understanding text, not images. Word-embedding and vector-search techniques are retrieval mechanisms that return existing content rather than create new visuals. Sequence-to-sequence speech models convert text to audio or vice versa and have no inherent image-generation capability. Therefore, a diffusion model best fits the need to create new marketing images from textual descriptions.