OpenAI's GPT Image 2, the company's first image model that reasons before it generates, became available inside ComfyUI on April 22, 2026 through the platform's Partner Nodes integration. The addition brings 2K resolution output, precise text rendering, and multi-image consistency directly into node-based workflows without any separate API setup.

What Happened

Comfy Org integrated GPT Image 2 (model ID: gpt-image-2) as an official Partner Node in ComfyUI v0.19.4. Unlike community-built custom nodes, Partner Nodes are maintained directly by the model provider and ship with first-party support. The integration is also available immediately on Comfy Cloud for users who do not run ComfyUI locally.

GPT Image 2 is described by OpenAI as a reasoning model: it performs an internal thinking step before generating an image, which produces more accurate handling of complex prompts and structured layouts.

Why It Matters

ComfyUI users gain access to a model specifically trained to handle two of the most common failure modes in AI image generation: legible text and precise editing. The model can generate images containing seven-item bulleted lists in a specified font, colorize photographs without warping faces, and maintain consistent character design across a batch of eight frames.

For studios and agencies running production pipelines on ComfyUI, these capabilities address real workflow gaps. Posters, social graphics, storyboard sequences, and product mockups all require accurate text and controlled edits — tasks where earlier diffusion models frequently fail. GPT Image 2 handling these inside the same node graph removes the round-trip to external tools.

The timing also connects to ComfyUI's $30M Series B announced two days later: the platform now controls integrations with both the open-source model ecosystem and closed frontier models from OpenAI, giving it leverage across the full professional AI creation stack.

Key Details

  • Model: GPT Image 2 (gpt-image-2), OpenAI's first reasoning-based image model
  • Access: Partner Node in ComfyUI v0.19.4 or Comfy Cloud
  • Resolution: Up to 2K output with structural integrity during edits
  • Text rendering: Handles complex layouts, specific fonts, and multi-element posters
  • Batch generation: Up to 8 images per prompt with consistent character and object continuity
  • Targeted editing: Modifies specified areas while preserving unedited regions pixel-accurately
  • Node name in library: "OpenAI GPT Image 1.5" (select gpt-image-2 model from the dropdown)

What to Do Next

To try GPT Image 2 in ComfyUI, update to v0.19.4 or later, open the Node Library, search for "OpenAI GPT Image 1.5", and select the gpt-image-2 model. You will need an OpenAI API key with image generation access. Comfy Cloud users can launch a pre-built workflow directly at cloud.comfy.org.

The full announcement with workflow examples is on the ComfyUI blog. If you are running video pipelines on ComfyUI, the Seedance 2.0 real-human video pipeline covers the video generation side of the platform's recent updates. The Comfy Org homepage has links to local install, cloud, and the GitHub repository.