openai / gpt-image-1.5 / createImage

Of the four OpenAI create nodes in this pack, this is the one to actually use. GPT Image 1.5 is OpenAI's current-generation text-to-image model - the December 2025 release that improved fidelity on top of the GPT Image 1 architecture - and CivitaiImageGenOpenaiGptImage15CreateImage is its ComfyUI face. Same cloud-execution story as the rest of the Civitai pack, but the model underneath is the one doing the heavy lifting that made GPT Image a household name in the first place: genuinely multimodal, excellent at rendering text inside images, and very good at following detailed natural-language prompts.

What it is. One of the ~160 generated nodes in civitai/civitai-comfy-nodes (engine openai, model gpt-image-1.5, operation createImage). Prompt in, IMAGE out, nothing local - the job runs on Civitai's fleet and you pay in Buzz. The whole appeal over calling OpenAI's API directly is that you skip the OpenAI account, key, and billing entirely; Civitai's auth and your Buzz balance are the whole story.

How it works. Standard pack machinery: the node builds a workflow payload, submits to orchestration.civitai.com, long-polls until the job lands, and returns four outputs - images (the IMAGE), errors, workflow_id, and raw_json (the full response, good for cost inspection). Required input is just prompt.

Inputs that matter. quality (low/medium/high, default high) is the main cost lever - drop to medium for drafts and watch your Buzz. size (1024x1024, 1536x1024, 1024x1536 - default 1024x1024) picks the canvas; the 1536-wide and 1536-tall options are where landscape and portrait compositions live. background (auto/transparent/opaque, default auto) can give you an alpha channel for assets. quantity generates multiple takes, and image_metadata (external metadata stored with the image) and output_format round out the optionals.

Getting good results. GPT Image reads full sentences better than any diffusion-model prompt convention - describe the scene, the light, the style, the text you want rendered ("a poster that reads 'COFFEE' in bold serif"). Its text rendering is the differentiator; lean into it.

Install. With the pack, once:

cd ComfyUI/custom_nodes
git clone https://github.com/civitai/civitai-comfy-nodes.git
pip install -r civitai-comfy-nodes/requirements.txt   # requests

Or ComfyUI Manager → Custom Nodes Manager → search Civitai Comfy Nodes (publisher civitai) → Install, then restart. Look under Civitai/Image/openai.

Auth and gotchas. Credentials resolve from a wired Civitai Auth node, then CIVITAI_API_TOKEN, then stored OAuth. Error hints: 402 = insufficient Buzz, 401 = token missing/expired, 429 = rate limited. Thirty-minute job cap, early-preview pack so shapes can shift. And the standard cloud asterisks: your prompt leaves the machine, and OpenAI's content policies - historically among the stricter ones - apply, with Civitai's moderation on top.

The take. If you want one reliable, current OpenAI create node in your graph, this is it. The 1.5's edit sibling is there when you need image input; for pure text-to-image with modern quality and legible in-image text, everything else in the OpenAI menu is either a step back (DALL-E) or a variant you may not need.