# Stable Dreamfusion Web UI
The Stable DIffusion Automatic1111 Web Ui is the primarily used web user interface for stable diffusion a 2d ai generation. Stable-Dreamfusion is a 3D txt to 3D and 2D to 3D ai mesh generator. Currently is only one poorly working and complicated google colab webui that doesn’t feature all of StableDreamfusion’s features online: https://colab.research.google.com/drive/1MXT3yfOFvO0ooKEfiUUvTKwUkrrlCHpF?usp=sharing

What we need is basically either an extension of the auto111 stable-diffusion webui that runs stable-dreamfusion or we need a new and separate gradio based web-ui for 3D mesh generation.

Below are the StableDreamfusion, StableDiffusion, and Zero 1 to 3 github links.

Do you believe this is possible.

https://vimeo.com/826423028?embedded=true&source=video_title&owner=200209865
https://github.com/ashawkey/stable-dreamfusion
https://github.com/cvlab-columbia/zero123
https://github.com/AUTOMATIC1111/stable-diffusion-webui
https://github.com/Mikubill/sd-webui-controlnet/discussions/204
https://gradio.app/docs/

There is also another route we can take. A Blender Plugin that runs a local (or online) copy of stable-diffusion. I have no long how long of an undertaking this would be BUT

This is the Shap-E blender addon

OpenAI Text to 3D Model Addon in Blender & BlenderText To 3D AI Generator

Here is my copy of the code you all to look at: https://drive.google.com/drive/folders/1rCGEvksJGRGI1VXdCwH4sdHCOprEJlSK?usp=sharing

The guy in the 8 min video said he made it in a day, it takes openai’s Shap-E https://github.com/openai/shap-e and runs the gpt3 (openai) api through Blender to run it. Now here is the problem, Shap-E sucks. It uses a point cloud method and not stable diffusion to generate object, and while it is useable, it’s really not ideal for the type of work we need to do.

There is a company called Kaedim that Julie and I were going to try before we came up with idea. https://www.kaedim3d.com/pricing The problem with them is that they are kind-of-not-really-but-yeah overcharging to run this pretty much free software through their website.

I think we can do better.

I’m sure the university has access to gpus that are greater or as good as GTX3090 so that our proposed “AutoMatt1111” web ui could be installed locally on a pc on campus or there are runpod and google collab implementations of automatic1111 web