
Hi guys,
I wanted to generate images from Copilot Studio, but I noticed this feature isn’t currently supported (it works in Copilot Studio Lite, though). So, I started researching alternatives using Azure AI Foundry and the Azure OpenAI Image API.
Here’s what I learned:
I can get OpenAI models to do Image Edits using an API directly in Azure, if you have an HuggingFace or Gemini API 🍌 you could use it
gpt-image-1 model.POSThttps://<your-resource>.openai.azure.com/openai/v1/images/editsAuthorization: Bearer <your-key>
image[] → File Content (from trigger or SharePoint)prompt → e.g., an icon of a beef steakmodel → gpt-image-1size → 1024x1024quality → mediumoutput_format → pngoutput_compression → 100n → 1Power Automate will handle the multipart boundaries automatically.
Some issues that I experiment
Content-Type: application/json. The API requires multipart/form-data.base64ToString()—send the Base64 content directly from Get file content.400 Bad Request, check that your headers and body format match the API requirements.