GPT img for ComfyUI

Languages: English | 한국어 | 中文

ComfyUI custom nodes for GPT image generation and ChatGPT LLM text responses through either OpenAI API keys or Codex/ChatGPT OAuth.

This node pack is ComfyUI-only. It does not run a separate web UI, Express API, gallery, session store, or local database.

Nodes

• GPT img OAuth Generate
• GPT img OAuth Generate Advanced
• GPT img OAuth Edit
• GPT img OAuth ChatGPT LLM
• GPT img API Generate
• GPT img API Generate Advanced
• GPT img API Edit
• GPT img API ChatGPT LLM

Image nodes return:

• image
• revised_prompt

ChatGPT LLM nodes return:

• text
• raw_response_json

Current Registry Status

As of 2026-04-25, this node pack is published to the Comfy Registry as:

node_id: gpt-img-node
publisher: hub2vu
version: 0.1.0
status: NodeVersionStatusPending
extract_status: success

This repository is now 0.1.6. Re-run the publish workflow to submit the newer version to the Registry. Because the published version is still Pending, it may not appear in ComfyUI Manager search yet. Manual Git installation works now. Manager installation should become available after a Registry version becomes NodeVersionStatusActive.

You can check the live Registry status with:

Invoke-RestMethod "https://api.comfy.org/nodes/gpt-img-node/versions?include_status_reason=true" | ConvertTo-Json -Depth 8

Manual Install

Clone this repository into ComfyUI/custom_nodes:

cd ComfyUI/custom_nodes
git clone https://github.com/hub2vu/Comfyui-GPT-img-node.git GPT-img

Restart ComfyUI after installing.

API Nodes

Use GPT img API Generate, GPT img API Edit, or GPT img API ChatGPT LLM.

Enter an OpenAI API key in the node's api_key field, or leave it empty and set OPENAI_API_KEY in your environment.

API usage is billed to the API key owner by OpenAI.

ChatGPT LLM Nodes

Use GPT img OAuth ChatGPT LLM or GPT img API ChatGPT LLM when you want a text response from a ChatGPT model inside a ComfyUI workflow.

ChatGPT LLM inputs:

• system_prompt: developer/system-style instructions for the assistant
• prompt: the user request
• model: the ChatGPT model to call
• reasoning_effort: minimal, low, medium, high, or xhigh
• max_output_tokens: maximum text output budget
• timeout_sec: request timeout

The API node also has api_key. The OAuth node also has oauth_port and auto_start_oauth.

ChatGPT LLM nodes also expose optional prompt input sockets:

• system_prompt_input: optional STRING socket that overrides system_prompt when connected
• user_prompt_input: optional STRING socket that overrides prompt when connected

If a prompt socket is not connected, the node uses the text entered directly in the node widget. The system_prompt widget is appended after the original image generation widgets so older saved workflows keep their existing widget order.

ChatGPT LLM model choices include gpt-5.5-pro. The image generation nodes keep their separate model list because pro models can have different streaming support.

Image Generate Prompt Inputs

Image generate nodes now expose both direct prompt widgets and optional prompt input sockets.

Prompt inputs:

• system_prompt: direct widget input for generation instructions
• prompt or advanced prompt fields: direct widget input for the user request
• system_prompt_input: optional STRING socket that overrides system_prompt when connected
• user_prompt_input: optional STRING socket that overrides the direct user prompt when connected

If a prompt socket is not connected, the node uses the text entered directly in the node widget.

Advanced Generate Nodes

Use GPT img OAuth Generate Advanced or GPT img API Generate Advanced when you want to separate the user's design request from reusable generation rules.

Advanced generate inputs:

• design_request: the authoritative garment/design request, in any language
• generation_instructions: reusable output format and quality instructions
• reference_instructions: how reference images should be interpreted
• hard_constraints: negative constraints such as no model, no logo, no text

The node automatically adds a conflict rule so design_request wins if another instruction conflicts with it.

OAuth Nodes

Use GPT img OAuth Generate, GPT img OAuth Edit, or GPT img OAuth ChatGPT LLM.

Requirements:

• Node.js with npx
• Codex/ChatGPT OAuth login

Login once if needed:

npx @openai/codex login

The OAuth nodes can auto-start the OAuth proxy on port 10531. Logs are written to:

custom_nodes/GPT-img/logs/openai-oauth.log

Notes

• No Python package install is required for this node pack.
• Do not commit API keys into workflows or repositories.
• The OAuth and API nodes use the same ComfyUI category: GPT img.