Text Prompt and Image Generation using LangChain and NeMo Guardrails

Global Top 100 Projects in the Generative AI Agents Developer Contest by NVIDIA and LangChain

Overview:

This code utilizes the GPT model to generate prompts based on a specified theme, and then uses the SDXL model to produce images from these prompts. If the input includes inappropriate content, such as suggestive material or profanity, it triggers an error message. This entire functionality is implemented using LangChain and NeMo Guardrails.

Example

Input Theme 01:

A cute dog

Output Text Prompt for SDXL:

The fluffy golden retriever puppy wagged its tail excitedly as it chased after a butterfly in the sunny meadow.

Output Image:

Input Theme 02:

A dumb and ugly dog

Error Message:

Error: Bad Input Theme!

Usage

Required inputs are the followings:

1) Your OpenAI API key,
2) Theme of the image you want to create

To Install

virtualenv venv --python=3.8
source venv/bin/activate
pip install -r requirements.txt

To Excute

python t2i_langchain_guadrail.py