PH's Archviz x AI Free ComfyUI Workflow

Dear Artists,

my name is Paul Hansen and i recently made a little tool for you with ComfyUI (https://github.com/comfyanonymous/ComfyUI, an outstanding node based opensource GUI for using generative AI models like StableDiffusion or FLUX locally on your computers), and like to share my latest work with you: a Workflow dedicated to Architectural Imagery made in ComfyUI involving StableDiffusionXL, Controlnet and FLUX. I've teased about it on reddit some weeks ago and recieved already lots of love from around the globe. By now it is available for free on https://civitai.com/models/920108/phs-archviz-x-ai-comfyui-workflow-sdxl-flux

It contains the examples shown in the showcase video on youtube https://youtu.be/6aXJqRhjXo0 to get you started, you will most likely be able to use it on your own projects in a very short amount of time, as there are many quality tutorials and examples on comfyui in general available.

The workflow is designed to enhance renderings or creates highres architectural images (tested 12288x8192) from lowres outputs of any kind of rendering software (tested 1536x1024) and tries to keep details throughout its process of "staged generation": a first stage with sdxl, a second stage for detailing first stages output with flux, a third stage upscaling the seconds stages output with flux again. If you have 3d people in your image, it will autodetect them and enhance them too. You can generate additional people with flux to your image by painting a mask. It brings a new level of control for generating imagery with controlnets through "detail conservation". It can be (and by its users already is) seen as a free alternative to many paid services, it although contains much knowledge usually kept behind paywalls. It is a project of my very own, no sponsoring or third party involved.

I use it mainly in the beginning and in the end of my process: i create previsualisations on the clients geometry to have a go-to reference for my 3d envirnoment very fast. After reaching a certain level of detail there, i enhance the outputs with some extra realism. This not only saves much energy i would have put into the first visuals after clay renderings, only then to hear that my vision is not what the clients was (yes, even discussing mood boards and example images in the first place - some of you may relate here). In the end i can at least use parts of the tools output for final images (like people, greenery, surroundings, furniture, you name it...). Considering a classic process without it, i can say that it saves me about 50% the time i usually spent on an image while getting enhanced quality and realism.

More information about, thoughts behind its creation and other insights awaits you on the initial reddit thread (and its comments): https://www.reddit.com/r/StableDiffusion

Attached images show examples of both main functionality: create an image from a low effort 3d scene (exterior file), enhance images (interior file)