StabilityAI Suite - Image Core + Style Preset
Stability's flagship hosted model in a node — with a style switch the author got merged upstream
- IMAGE
Image Core is the front door to Stability AI's hosted flagship - the model behind stable-image/generate/core. It's the API-first SD3.5-class model, historically the one that ran circles around the open 2B SD3 Medium that everyone actually got to download. You can't run this thing locally; that's the whole point of the node. It's also why you need an API key and credits.
The interesting part of this fork: the author updated the official Image Core node with a true/false switch for styles plus a style-preset selector, then submitted the change upstream - where it was accepted. So this node's party trick eventually landed in the official Stability node too. You're getting the enhanced version here.
How it works
A straightforward POST to stable-image/generate/core with your prompt, an aspect ratio, and (optionally) a style preset. Unlike Creative Upscale there's no polling - the image comes straight back in the response, so it's fast and simple. When style is off (the default), the node doesn't even send the preset.
Inputs that matter
prompt- the only required input.aspect_ratio- nine options from 21:9 down to 9:16. The easy knob for composition.style+style_preset- flip the switch on, then pick from 17 presets: 3d-model, analog-film, anime, cinematic, comic-book, digital-art, enhance, fantasy-art, isometric, line-art, low-poly, modeling-compound, neon-punk, origami, photographic, pixel-art, tile-texture. Great for keeping a set of images visually consistent without prompt engineering.
negative_prompt, seed, and output_format (png/webp/jpeg) round it out. Output is a single IMAGE.
Install
The standard pack install. ComfyUI Manager → search "ComfyUI StabilityAI Suite". Or:
cd ComfyUI/custom_nodes
git clone https://github.com/florestefano1975/ComfyUI-StabilityAI-Suite
Restart, then set the key - SAI_API_KEY env var, sai_platform_key.txt, or the api_key_override field on the node. Nothing to download; this pack ships no models and needs no extra pip packages.
Gotchas
- Paid per image, same as every node in the suite.
- Style presets are off by default; if you toggle the switch off mid-workflow the node silently drops the preset entirely.
unauthorizedmeans the key;payment_requiredmeans the balance.
One honest note: this is a convenience node, not a magic one. For batch or serious work, most people still prefer local SDXL or Flux where it's free after the GPU cost. Where Core shines is one-off quality with zero local setup - paste a prompt, get a polished SD3.5-class result, and never babysit a multi-gigabyte model download.
Inputs (8)
| Name | Type | Default | Description |
|---|---|---|---|
| prompt | STRING | — | |
| negative_promptopt | STRING | — | |
| seedopt | INT | 00–4294967294 | — |
| output_formatopt | COMBO | 3 options: png, webp, jpeg | |
| aspect_ratioopt | COMBO | 9 options: 16:9, 1:1, 21:9, 2:3, 3:2, 4:5, +3 | |
| styleopt | BOOLEAN | false | — |
| style_presetopt | COMBO | 17 options: 3d-model, analog-film, anime, cinematic, comic-book, digital-art, +11 | |
| api_key_overrideopt | STRING | — |
Outputs (1)
| Name | Type | Description |
|---|---|---|
| IMAGE | IMAGE | — |