Architects' tools

Now go to ComfyUI. actually workflow I've already posted yesterday, here is just a different prompt and some parameters. Make sure you have everything installed as specified here. I use this workflow very often to find conceptual ideas, modifying Checkpoint, Lora, sampler...

With Vary(Strong), in turn, you can get interesting results with good detail, but very different from what was in version 5.2, which is also great in our case, because the result from version 5.2 we can already use in ComfyUI and if we have more material, it's only better.

Vary(Subtle) result is closer to the original, but usually gives a lot of artifacts, which for our purposes, however, is not a big problem, since we will still change these images in ComfyUI.

By the way, this prompt structure does not work very well with midjourney 6.0, but you can try it, as this version perfectly copes with details and with creating realistic images. There is another trick that allows you to use version 5.2 and 6.0 in combination. First you need to enable Remix mode in settings, which gives you the ability to change the prompt when creating variations of the image. Next we will just try to make such variations, but already in version 6.0. We have two options for this Vary(Subtle) and Vary(Strong). Choose one of them and just change the version in the corresponding parameter.

as an example let's take this image, I generated it in midjourney 5.2 as this version often gives more interesting results of the shape of a building or object compared to version 6.0. prompt was like this: [ birds nest with dozens of swings] ::1 + [ black voxelized ceramic 3d printed bridge in Amsterdam with dozens of swings] ::1.5 + award winning architectural photo ::1 --no birds --ar 3:2 --v 5.2 this prompt contains two significant parts, I conventionally call them idea and object. The idea is usually a description of some natural structure or pattern that I want to mix with the architecture, in this case it is written next way: [ birds nest with dozens of swings] object is a direct description of an architectural object, its materials and properties: [ black voxelized ceramic 3d printed bridge in Amsterdam with dozens of swings] after brackets, double colons indicate weights, they are necessary to regulate the degree of mixing of the idea with the object.

I used images generated in #midjourney as inputs for the comfyui workflow, here are some of them

These images are produced using the same workflow in #comfyui , but other programs were also used. There will be several posts today, I will tell you exactly how it is done in all stages.

workflow file and input image #comfyui

I built a workflow in #comfyui , which preserves the composition of the original image, but changes the details and style of the main object in the frame. in this example, the origami style is selected (you can change it to another throught the text prompt). what is used here: 1. TTPLanet_SDXL_Controlnet_Tile_Realistic (my workflow was built based on the example that can be found in the description of this model on civitai) 2. Prompt Styler 3. LORA FormFinder-XL, which was trained by Ismail Seleit 4. RealVisXL model the post shows pairs of images, original which i made with midjourney + result after rendering in comfyui

Of course, you can use any parts and any volumes for your structures.

This example is more complicated, there are more parts and the connection rules are defined by the algorithm

Files of this and the following example #ghsample

The first example is quite simple, it uses 5 different parts and the rules are set manually

Generating structures by aggregation. Grasshopper has two main plugins for this task: Wasp and Monoceros. Wasp creates structures that grow depending on the number of iterations. In Monoceros, the structure fills a predefined volume. In both you have to prepare the parts beforehand and explain to the program the rules of connecting these parts. Today we will talk about Monoceros. #ghplugin

the algorithm isn't complex at all