/0.0.2

0.0.2 | Stable Diffusion Hackathon

Primary LanguageJupyter NotebookMIT LicenseMIT

0.0.2

0.0.2 | Stable Diffusion Hackathon

Challenge

Using AI Image Generation tools, build an avatar builder that allows the user to create a customizable avatar using the least amount of prompts to the user. Feel free to build anything to accomplish. You can run stable diffusion on your box and build a custom UI on top of it or just use online tools for image generation.

Submit your solution to this repo in a folder with your team name.

Deliverables:

  1. AI Prompt (i.e. Mad libs for ChatGPT)
  2. Backend code running image generation on your box (optional)
  3. User Interface (optional)
Example Models:

RunwayML's original Stable Diffusion1.5 model

RunwayML Example

Stable Diffusion 2 - StabilityAI's model for 512x512

StabilityAI example

Midjourney Replica

Midjourney Example

SDXL - Generates in 1024x1024, a little more difficult to set up but could make better results

SDXL VAE

SDXL Refiner

SDXL Refiner VAE

Apple's SDXL for running on Mac M1/M2 GPUs

SDXL Example

3D Animation Diffusion base model - Disney/Pixar style outputs

3D Animation Example

AbsoluteReality - Realism focused model

Absolute Reality

Dreamshaper8 - Distinctly AI generated model

Dreamshaper Example

Example Prompt:

Chibi-style, 3D, <Descriptor> student using a backpack, <clothing>, holding a <item>, wearing headphones, is a <descriptor>

Filled In Example:

Chibi-style, 3D, Swiss student using a backpack, lederhosen, holding a laptop, wearing headphones, is a boy

Generated Image

Example Avatar