MELDI for Amore Seongsu
Summary
As the Graphic and VMD Manager at Amore Seongsu, I developed characters and the corresponding visual assets needed for brand-exclusive content communication using AI tools.
Rather than relying on a single AI platform, I strategically combined the strengths of multiple AI tools to achieve the desired creative outcome.
The process of building a world and developing ideas typically requires significant time; however, by leveraging Chat GPT during the early stages, I was able to quickly construct a coherent world view and storyline.
Based on this established narrative and visual concept, I then used Stable Diffusion to generate background and character images separately.
Finally, the most suitable background and character elements were composited in Photoshop to produce the finished visuals.
Using Stable Diffusion
Changing the Seasons in the Garden
Changing the Seasons in the Garden
Step 1. Find a background image
Select a landscape image of the Amore Seongsu Central Garden to use as the background.
Select a landscape image of the Amore Seongsu Central Garden to use as the background.
Step 2. Spatial recognition with ControlNet in Stable Diffusion
Apply ControlNet using the ‘LineArt’ control option to recognize the spatial structure of the image.
Apply ControlNet using the ‘LineArt’ control option to recognize the spatial structure of the image.
Step 3. Generate a new image using prompts in Stable Diffusion’s ControlNet
Prompt: winter garden, lights on trees, Christmas lights, warm light, snow, cozy
Prompt: winter garden, lights on trees, Christmas lights, warm light, snow, cozy
Convert the generated image into a video using Runway AI.
Review
At the early stage, I explored Midjourney and Stable Diffusion to create a new character exclusive to Amore Seongsu.
, since these tools rely on vast amounts of pre-trained data, it proved challenging to generate truly original and beautiful characters that do not already exist.
Moreover, rather than being passively guided by AI’s output, I found that directly using Photoshop or complementary AI tools for areas beyond AI’s capability not only saved time but also significantly improved the quality of the final results.
In the case of Stable Diffusion, experimenting with various checkpoints to find the optimal visual style was an essential part of the process.
Satisfying points
Because the background was based on the real Amore Seongsu space, it was necessary to generate images that closely resembled real-world environments.
Stable Diffusion provided the perfect solution for this — it effectively incorporated new visual elements into the existing space, accurately reflecting the atmosphere based on the given prompts.
Disappointing points
Consistency in character depiction remains a key challenge in AI-generated imagery.
Even when entering the same prompt, the resulting images often vary slightly, which highlights one of the current limitations of AI tools.
- Amorepacific Creatives
- Planning & Design
- 신정식
