Recently, we conducted the first phase of the enhancement work on WEINOVEL's image generation system.

The existing block-based image generation method only allowed the AI to analyze the content of that block, while the rest of the scene description had to be input directly by the author.

However, in the actual creative process, this was not as easy as it seemed.

The author had to describe the scene without knowing how much context the AI could understand, and if the description was insufficient, images were generated that did not match the intent. Conversely, if the description was too lengthy, it became burdensome to write the prompt.

Therefore, in this enhancement work, we completely changed our approach.

Now, the AI does not simply look at the current block but first understands the entire content of the episode. It analyzes what role the scene plays within the episode and what meaning it has in the overall flow of the novel before generating the image.

In simple terms, the core goal of this work was to make the AI not just a simple image generator but to perform the role of an 'image director.'

The second area of focus was the diversity of scene direction.

What WEINOVEL pursues is not just creating pretty illustrations but images that convey a story.

Therefore, we improved the AI to directly choose shooting composition, focus, and camera angles to express relationships between characters, emotional flow, and tension in the scene.

We are conducting research to utilize various directing techniques such as close-ups, over-the-shoulder shots, wide shots, and leading lines to generate images that fit the flow of the story, even for the same scene.

Of course, there are still many challenges to be addressed.

Complex scenes with many characters, detailed emotional expressions, and maintaining consistency between scenes are areas that will need continuous improvement.

Nevertheless, I believe that through this first phase of enhancement work, WEINOVEL has reached a stage where it can capture much more diverse stories in images than before.

We will continue to develop towards the goal of "images for stories" in the future.