PD: GPT Image 1 & 1.5 (comfyui_AuthToken)
GPT Image 1 & 1.5 via session token — the no-new-key option
- image
- mask
- image
- info
Every image model in the comfyui-PD_comfy-api-node pack gets an API-key version and an AuthToken version, and GPT Image 1 & 1.5 is no exception. This is the token flavor: same engine as PD: GPT Image 1 & 1.5 (apikey) - GPT Image 1 and the December 2025 1.5 refresh, run through ComfyUI's hosted proxy - but you authenticate with a session token instead of a minted ComfyUI API key.
When to pick this over the key version
The token is the auth_token_comfy_org JWT from your logged-in ComfyUI session, grabbed with the pack's bundled Tampermonkey userscript at temperay/PD_comfyui_token_grabber.user.js. Load the app, log in, let the script lift the token out of localStorage, paste it in. The node sends it as Authorization: Bearer <token>.
It's the lower-friction path if you already have a ComfyUI account and don't want to manage a key - and the higher-fragility path, because session tokens expire and carry your full account permissions. For a quick concept image tonight, it's great. For a workflow you'll rerun next month, the key variant is the one that won't 401 on you mid-run.
Inputs and outputs
These match the API-key twin field for field:
- auth_token - the grabbed session token.
- prompt - what to generate or edit.
- model -
gpt-image-1orgpt-image-1.5(default). - quality -
low/medium/high(default low; drives cost and time). - background -
auto/opaque/transparent- GPT Image 1 still has transparent-background support, unlike GPT Image 2. - size -
auto,1024x1024,1024x1536,1536x1024. - n - 1–8 images.
- seed - present here with a full range; still treat results as non-deterministic.
- Optional image (editing) and mask (inpainting, white = replaced).
Outputs: image and info.
Installing and gotchas
cd ComfyUI/custom_nodes
git clone https://github.com/7BEII/comfyui-PD_comfy-api-node.git
Restart, find it under PD_Tools/Image_Generation. (The README's pip install -r requirements.txt is a dead instruction - no requirements file ships; the pack's real dependency is ComfyUI's official API nodes, bundled with Desktop.)
Troubleshooting is token-shaped:
- 401/403 - the token expired. Re-run the grabber.
- Black output - error placeholder;
infohas the message. - Transparent backgrounds flaky - test one call before depending on it, since the proxy path can vary.
- Don't share workflows with a token baked in - it's a live credential.
If token expiry is starting to annoy you, the sibling API-key node gives you the same GPT Image 1/1.5 engine with a credential that doesn't rot - and if you want the newer model with multi-image editing, that's GPT Image 2, also in this pack.
Inputs (10)
| Name | Type | Default | Description |
|---|---|---|---|
| auth_token | STRING | — | |
| prompt | STRING | Describe what you want to generate or edit | |
| model | COMBO | gpt-image-1.5 | GPT Image model version |
| quality | COMBO | low | Image quality, affects cost and generation time |
| background | COMBO | auto | Background transparency |
| size | COMBO | auto | Image size |
| n | INT | 11–8 | Number of images to generate |
| seed | INT | 00–18446744073709550000 | — |
| imageopt | IMAGE | Input image for editing | |
| maskopt | MASK | Mask for inpainting |
Outputs (2)
| Name | Type | Description |
|---|---|---|
| image | IMAGE | — |
| info | STRING | — |