Ideogram V4.5 Image-to-Image: Realistic and Artistic Portrait Dual-Style Demonstration

This entry uses the same portrait source image to demonstrate the differences of Ideogram V4.5 image-to-image under two stylistic directions: one side preserves realism (natural skin texture, soft window light, restrained color grading), while the other takes an artistic approach (painterly brushstrokes, stylized lighting). Both versions control direction only through the prompt and the quality tier, with the source image as the sole conditional input, making it easy for developers to compare the output distribution of "same input, different prompts." At the schema level, image_url is the required primary source image, bound after generation by the pipeline; quality is set to high to observe the fidelity of the high-quality tier on skin texture and edges; image_size=portrait_4_3 fits the portrait composition; num_images is fixed at 1. mask_url and reference_image_urls are not used, keeping the minimal input surface for single-source-image editing. This is suitable for scenarios that need to establish image-to-image baseline samples: building a prompt template for the team, evaluating the boundary between realism and stylization, and validating the first-step rewrite effect before multi-turn editing. According to public information, this model emphasizes precise editing and stability across multi-turn editing, so this baseline can serve as a preliminary step for subsequent local replacement.