I used ChatGPT Images 1.5 and Gemini Nano Banana Pro to turn my selfie into K-pop and Chibi art—here’s what happened

OpenAI has proclaimed the world-wide rollout of ChatGPT Images 1.5 on 17 dec, offering a john major rising slope to its AI image generation and editing capabilities. The new version, powered by the flagship GPT Image 1.5 model, is claimed to offer faster, more precise, and flexible image creation and editing directly within ChatGPT. Interestingly, the AI model challenges Google’s Gemini Nano Banana Pro, which received major updates in August. Users and developers worldwide can now access these enhanced visual tools through ChatGPT and its API.

In this article, we compare the results of both the AI tools, as two similar prompts were used to generate the outputs.

Comparing AI performance with common prompts

To test the upgrades, two prompts were used across both ChatGPT Images 1.5 and Gemini Nano Banana Pro.

Prompt 1 (K-pop Idol Transformation)
Using the provided image of the subject as a reference, transform them into a K-pop idol–style version of themselves, fully preserving their natural facial features, skin tone, ethnicity, and identity. Style the subject with a polished, high-fashion idol aesthetic inspired by contemporary K-pop concept photoshoots, featuring editorial studio lighting with a soft glow and clean highlights, a flawless yet natural dewy skin finish, and subtle enhancements to the eyes, lips, and hair for a camera-ready look. The subject poses confidently with expressive but controlled body language, styled in fashion-forward outfits influenced by modern K-pop trends such as elevated streetwear, Y2K accents, chic tailoring, glam punk, or soft ethereal looks, adapted to complement their original clothing style. The atmosphere resembles a professional idol photoshoot, incorporating bold coloured backdrops or moody dramatic environments, studio or concert-style lighting, cinematic shadows, and refined colour grading, with optional tasteful details like layered jewellery, belts, or statement accessories kept cohesive and restrained. The final image should feel like an authentic K-pop concept photo—crisp, stylish, and aspirational—projecting polished charisma and star presence while clearly remaining the same person.

Both AI tools performed exceptionally well. While Gemini generated images faster, ChatGPT Images 1.5 produced more vibrant outfits and backgrounds, though it took approximately 60 seconds to render compared with Gemini’s 35–40 seconds.

Prompt 2 (Chibi Character Transformation)
Prompt: Transform the subjects or image into an adorable chibi-style character with a tiny body and an oversized head. If the image contains a person or multiple people, give them large, sparkling eyes, soft rounded facial features, and a cheerful expression, while preserving their recognisable traits such as key facial features, hairstyle, accessories, or distinctive clothing. If the image contains an object, animal, or scene, reinterpret its most recognisable features using the same chibi proportions and simplified, cute styling. Keep the overall look short and cute, with smooth pastel shading and simplified details. Make the final image bright, expressive, and irresistibly charming, like a collectible chibi mascot.

For chibi transformations, both ChatGPT and Gemini produced high-quality results, though Gemini captured background elements slightly better and rendered more realistic chibi facial and clothing features. Gemini also completed the images faster (around 40 seconds) than ChatGPT (around 60 seconds).

Conclusion

ChatGPT Images 1.5 represents a strong step forward in AI-powered image generation, emphasising vibrant visuals, precise editing, and flexibility. While Google’s Gemini Nano Banana Pro still offers faster performance, OpenAI’s update is a clear attempt to match, if not surpass, its rival in creative control and output quality, particularly for detailed and styled transformations.

Enhanced image editing features

The standout feature of ChatGPT Images 1.5 is its ability to edit only selected parts of an image while keeping the rest untouched. Users can remove or add objects, change colours, or adjust styles without compromising the original look. The model also supports combining multiple images into one cohesive scene, giving users creative control over complex compositions.

One of the major improvements lies in instruction-following. When users provide detailed editing instructions, the AI changes only what is requested, ensuring:

• More realistic photo edits
• Believable clothing and hairstyle try-ons
• Cleaner stylistic filters and conceptual transformations
  
  The model’s advanced capabilities include adding, removing, blending, combining, and transposing elements without degrading image quality.

---

---