Griptape Driver: Azure OpenAI Image Generation

If your DALL-E credits flow through Azure OpenAI rather than OpenAI directly - a common arrangement for anyone with an enterprise Azure contract - then the plain OpenAI image driver won't work for you. Azure wraps the same models behind your own endpoint and deployment names. Griptape Driver: Azure OpenAI Image Generation is the version that speaks that dialect.

It's a driver you reach for for one reason: your organization's keys and billing are on Azure. If you're an individual with a normal OpenAI account, skip it and use the OpenAI driver instead - Azure adds deployment-management overhead without changing the images.

How it works

Configuration node again: the DRIVER output plugs into a Griptape Run: Prompt Image Generation Task, which calls Azure's image endpoint with your prompt and returns the image. The node itself just builds the driver.

The inputs that matter:

• image_generation_model - dall-e-3 or dall-e-2. The image model, not the deployment.
• image_deployment_name - the deployment name you created in the Azure portal for that model. This is the Azure-specific bit; you can name a deployment anything, and this field has to match yours, not a default.
• size - 256, 512, or 1024 square, plus 1024x1792 / 1792x1024 portrait and landscape for DALL-E 3.

Then the credentials, with the pack's usual naming twist: image_endpoint_env_var and image_api_key_env_var expect the names of environment variables (AZURE_OPENAI_DALL_E_3_ENDPOINT, AZURE_OPENAI_DALL_E_3_API_KEY by default), not the endpoint and key typed in directly. The README links Microsoft's guide for getting those values.

Installing

Ships in the ComfyUI Griptape Nodes pack:

• ComfyUI Manager: search "Griptape" → install ComfyUI-Griptape.
• Manual: cd ComfyUI/custom_nodes && git clone https://github.com/griptape-ai/ComfyUI-Griptape, then restart.

Pack dependencies: griptape[all], openai, python-dotenv, plus git-hosted extensions. The pack-wide torch caveat applies - Griptape installing torch can mismatch ComfyUI's build; the README's fix is a torch reinstall with --extra-index-url https://download.pytorch.org/whl/cu121.

Gotchas

The classic Azure failure is a deployment mismatch: the default deployment name is dall-e-3, but if yours is called dalle-prod or whatever, you must set it or you'll get 404/401-style errors that look like key problems but aren't. DALL-E 3 on Azure also silently enforces content policies - prompts that sail through local SD will get refused, often with a generic error. And if your env vars are already set system-wide, the pack's Settings → Griptape screen shows them, so don't type them into node fields; the node wants variable names, not secrets.