So, the next step was to create 3D models. For tasks like this during the concept design stage, I always use #rhino, and this time was no exception. There are some screenshots in this post, which I used as inputs for renders.

Today’s renderings. The screenshots from Rhino shown in the previous post were used as the basis for them + I've added as inputs a few textures for materials references. Everything was fed into the Nano Banana 2.0 node in #comfyui, where i made multiple generations for each view in 2K resolution. Then, a collage was assembled in Photoshop from the best parts of each generation (for example, in one generation the building itself turned out just right, while in another the textures of the sea and sand were perfect - these were combined to one image, but in reality from smaller pieces); on the next step the image was sent back to ComfyUI for upscaling to 4K (lately I’ve been using the Magnific node for this task); and finally, postprocessing in Photoshop (color and lighting correction, and regenerating small details and people for better quality). It usually takes me 2–4 hours to complete a single render, from the screenshot to the final result.

Some of the most interesting images in this project - in terms of working on them - are the orthogonal top views, which were used as explanatory diagrams. First, a close-up render was created where only the piers without much of the surrounding, and then that render, along with a screenshot from Rhino showing color codes, was returned as an input to #comfyui to generate the second much more complex render. Of course, Nano Banana can’t perfectly interpret and render every color code on the first try, but that’s not the goal: if even one color code turns out well, that’s good enough. Then, as I described in my previous post, it’s collage > upscaling > inpainting for improving details.

The images in this post are in the order they appeared in the workflow. And here’s the prompt, in case anyone’s interested:

IMPORTANT: PRECISELY FOLLOW WHITE CONTOURS AND KEEP SIZE OF ALL ELEMENTS FROM THE ATTACHED IMAGE. KEEP CAMERA VIEW, PROPORTIONS, SCALE AND ASPECT RATIO OF THE ATTACHED IMAGE.

for the first attached image make an Ultra-realistic architectural visualization with the next features:

- use the second attached image for all the materials and surfaces as reference
- colorful area in the middle of the first attached image is the park divided by zones which are corresponded to the colors:
1. light blue color is represented water
2. yellow color is represented beach
3. light brown colored zone is represented a biodiversity park organized with a grid structure where each grid cell is a different plants or surface as shown on a top side of the second attached image
4. grey colored zone is represented a public square primarily with a hard surfaces
5. rose area is represented a market
6. light green is represented a market picnic zone with a grass and palm trees islands
7. dark green is represented a city farm zone
8. red is represented a sport zone
8. dark blue is represented a commercial and touristic information zone


Style & Rendering Quality:
- Summer afternoon
- Modern district is Hainan, China
- Photoreal lighting, global illumination. low contrast early morning shot.

Limits:
- no cars

Another pier from this project (there are three in total). This one is intended solely for swimming and recreational activities, with no market area. It will be built on the site of the existing pier. A second swimming area is needed, first, because it is far from the first one (the distance between the piers is approximately 1.5 km), and, second, because swimming in this bay, despite the good water quality and the beach, is only possible in designated areas due to strong rip currents; you simply cannot enter the water anywhere along the beach - special areas with nets and lifeguards are required. In this situation, the piers help create such areas and serve as attractions.