There are several ways to accomplish the image-to-image pattern. Here I explore the following flow where a caption is extracted from the source image and then the caption is injected into a prompt to generate the target image:

• 🔍 Bing Search retrieves images based on a user query (this step may be replaced with a custom image upload)
• 👁️ Azure AI Vision analyzes the image and detects recognizable attributes
• 🎨 DALL·E 2 from Azure OpenAI generates a new image with a prompt composed by the detected attributes and instructions such as clipart

• Node.js 8 or later
• A valid subscription key for the Bing Search APIs
• A valid subscription key for the Azure AI Vision APIs
• A valid subscription key for the Azure OpenAI APIs with DALL·E 2 enabled

1. Clone the repository.

2. Navigate to the cloned directory.

3. Install Express.js: npm install

4. Run the sample app: node app.js

5. Navigate to the provided URL and have fun doing image-to-image with Bing Search, AI Vision, and DALL·E 2!