Olivio Sarikas begins with loose character prompts for cute monsters, fantasy creatures, and a monkey monk with a pet mouse. The experiments show that broad concepts can quickly produce visually appealing directions, but the useful result depends on whether clothing, hands, props, anatomy, and repeated motifs remain logically consistent when inspected closely.
His side-by-side tests favor ChatGPT for detailed character sheets and coherent accessories. It adds expression studies, color palettes, names, personality notes, and props that support a character's role. Midjourney often delivers stronger mood and striking silhouettes, but Sarikas repeatedly finds floating weapons, malformed hands, decorative details that do not resolve, and objects that only resemble the requested item from a distance.
The stream also demonstrates a broader workflow: use one model to explore concepts, provide references or sketches when text cannot describe a novel form, repair weak details with another model, and animate a selected still with a video generator. Sarikas concludes that tool choice should follow the production goal, with coherent detail and controllability mattering more than an attractive first impression.
Watch on YouTube
