Exploring Image Generation with Gemini 2.0: Insights from My Recent Experiment

Recently, I delved into the capabilities of Gemini 2.0 for image generation and decided to embark on creating an entire manga one-shot. Inspired by a post showcasing a detailed manga panel, I was eager to explore what this advanced tool could achieve. However, after spending a couple of hours on this task, I made a few notable observations.

First and foremost, the initial page generation displayed impressive quality, capturing the aesthetic I’m aiming for. Unfortunately, I noticed a decline in quality with subsequent pages. This degradation is something to consider for anyone aiming for a cohesive visual narrative across multiple panels.

Another challenge I faced was character consistency. Throughout the process, the generated images often strayed from the intended design. Whether it was the characters’ clothing appearing differently from one panel to the next or inconsistencies in the environment, maintaining a uniform look became a struggle.

Despite these hurdles, I’m sharing my best-generated page, which I believe showcases the potential of Gemini 2.0. Engaging with this technology is an exciting journey, and I’m looking forward to refining my approach and achieving better results in future projects. Stay tuned for more updates as I continue to explore the world of digital imagery!