GPT-Image-2 ComfyUI
https://github.com/Anil-matcha/gpt-image-2-comfyui- Category
- AI Tools
- Rank
- No. 1712Tools index
Previous survey · No. 1718 ·
- Pricing
- Open Source
- Type
- TOOL
- Builder
- anil-matcha
- GitHub
- 2 stars
- Date
About
ComfyUI custom nodes for GPT-Image-2 generation via muapi.ai — text-to-image and image-to-image.
What it does
It turns remote image generation and editing into composable ComfyUI graph steps. Users supply a prompt, optionally connect up to nine reference images, then receive the resulting image tensor alongside its hosted URL and request identifier.
Why it's ranked here
This is a focused integration with useful workflow ergonomics. It handles authentication, reference-image uploads, job polling, result downloads, and ComfyUI tensor conversion. Its narrow scope and reliance on one hosted service make it practical for existing ComfyUI users, but unsuitable as a general image stack.
What's good
The three-node design separates credentials from generation, so one key node can serve multiple graph branches. Image editing accepts as many as nine references. Both generation nodes return a native image tensor, hosted URL, and request identifier. Included example workflows demonstrate preview-ready text and editing graphs.
Tradeoffs
Every generation request depends on muapi.ai, an API key, available credits, and network access. Jobs poll every five seconds and stop after ten minutes. Reference tensors are converted to JPEG before upload, which is lossy. The exposed seed only defeats ComfyUI caching and is not sent as generation control.
How to use it well
Choose it when ComfyUI is already your orchestration layer and you want hosted prompt generation or multi-reference edits inside larger visual graphs. Configure credentials once, reuse that connection, and retain returned request identifiers for tracing. Look elsewhere for local inference, offline work, deterministic seed control, or a standalone application.
Technical notes+
gpt_image2_nodes.py defines GPTImage2ApiKey, GPTImage2TextToImage, and GPTImage2ImageToImage, registered through NODE_CLASS_MAPPINGS and exported by __init__.py. Requests authenticate with the x-api-key header. Image inputs are converted from Torch tensors to quality-95 JPEGs and uploaded before submission. _poll checks status every five seconds for up to 600 seconds, while _download_image converts the returned RGB image into a batched float tensor. requirements.txt declares requests, Pillow, NumPy, and Torch. GPTImage2_T2I_Example.json and GPTImage2_I2I_Example.json provide importable workflow examples.
Observed
- License
- MIT License
- Primary language
- Python
- Install surface
- ComfyUI custom node, installable through ComfyUI Manager by Git URL or manually with pip requirements
- Interface
- ComfyUI nodes backed by the muapi.ai HTTP API
- Dependencies
- requests, Pillow, NumPy, and Torch
- Authentication
- API key through a node input or the MuAPI CLI configuration file
- Included examples
- Separate JSON workflows for text-to-image and image-to-image
Read from README.md, requirements.txt, __init__.py, gpt_image2_nodes.py, LICENSE, GPTImage2_I2I_Example.json, GPTImage2_T2I_Example.json.
What it can do
Generate images from text descriptions
Text prompt describing desired image → Generated image
Transform existing images based on text prompts
Source image and text prompt for modifications → Modified image
Create custom ComfyUI workflows for GPT-Image-2
Node configurations and parameters → ComfyUI workflow for image generation
Connect to muapi.ai service for image processing
API credentials and generation requests → Processed images via muapi.ai
Integrate GPT-Image-2 capabilities into ComfyUI interface
ComfyUI installation and node setup → Extended ComfyUI with GPT-Image-2 nodes
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.