Text to Image using Voiceflow API Step and Hugging Face

Overview

This repository contains a Node.js application with an endpoint /text2image leveraging Hugging Face's textToImage function of their inference API.

Tutorial Video

To aid in understanding the application code and constructing a Voiceflow project using it, a detailed Youtube video tutorial is available at this link.

The tutorial provides a walkthrough of the code functionality, as well as a practical guide on integrating it with a Voiceflow project for testing and use.

How to Use

To use this application, make a POST request to the /text2image endpoint with a JSON body consisting of the following parameters:

• token (optional): Your Hugging Face token.
• model (optional): The model used for the text-to-image function. If not provided, it defaults to 'promptero/openjourney-v4'.
• prompt: The text prompt to convert into an image.
• parameters (optional): The parameters to customize the model inference.

For example:

{
  "prompt": "YOUR_PROMPT_TEXT"
}

{
  "token": "YOUR_HUGGINGFACE_TOKEN",
  "model": "promptero/openjourney-v4",
  "prompt": "YOUR_PROMPT_TEXT",
  "parameters": { "OPTIONAL_PARAMETERS" }
}

Note: Please ensure to replace YOUR_HUGGINGFACE_TOKEN, YOUR_PROMPT_TEXT, and optionally provide parameters according to your needs.

Return format:

The API returns a JSON object with a buffer attribute as an image (Base64 format string).

Failure to provide a prompt will result in an HTTP 400 response with the error message: "Missing required parameters: prompt".

In case the image generation fails, the API will return an HTTP 500 response with an error message.

Testing the API using ngrok and Voiceflow API Step

This created endpoint can be easily tested and used with Voiceflow, by first exposing your local server to the internet using ngrok, and then calling the resulting URL in Voiceflow's API Step.

To expose your local server using ngrok:

• Download and install ngrok
• Run your server (The port is 3800 in our current application)
• Expose your server to the web using the command ./ngrok http 3800
• Copy the Forwarding URL from the ngrok command line interface

Now in the Voiceflow API Step, paste the URL copied from ngrok followed by the /text2image endpoint, and specify the method as POST. In the Body section, add the required parameters as mentioned in the "How to Use" section.

Contribution

You're welcome to contribute to this project. Please fork the repo, make your improvements and then submit a PR.