In recent years, significant advancements have been made in the field of text-to-image generation using artificial intelligence. Research teams have proposed a method for training open diffusion models using Creative Commons images. By employing transfer learning techniques, they match synthesized captions with Creative Commons images, thereby transforming text into images. This approach has substantially improved the quality of generative models, creating a set of models known as CommonCanvas that rival SD2. Through this method, they have overcome the limitations of dataset size and the use of artificial captions, achieving high-quality image generation. Overall, their research provides new methods for utilizing Creative Commons images and offers robust support for the further development of generative models.