Stability Image Core

Image Core is the plain-text-to-image node of Stability's official ComfyUI pack - the one you reach for when you just want a good image from a prompt and you want Stability's servers to do all the work. It hits the stable-image/generate/core route, and "core" is the name for Stability's mid-tier generation model: better than the legacy SD3 route, a step below the flagship Image Ultra, and the sensible default for everyday generation.

Why would you use this when local Flux and SDXL exist and are free? Same answer as every node in this pack: no VRAM, no model management, no downloads - just a prompt and a paid API call. The community's take on this pack has never recovered from the launch thread's "Not local, not interested," and honestly, if you have a working local stack, you don't need this. Where it earns its keep: machines without a decent GPU, quick experiments while you're away from your rig, and any workflow where you want Stability's own model's look without babysitting weights.

How it works

One required input: prompt. The rest is what you'd expect from any hosted generator:

• aspect_ratio - nine presets from 1:1 up to 21:9, and the dropdown labels actually carry the target resolution in parentheses (1:1 renders at 1536×1536, 16:9 at 2016×1152). The node strips the resolution before calling the API; you just pick the shape.
• negative_prompt - for steering away from things.
• seed - set it, reproduce it.
• style / style_preset - pick from Stability's preset list (anime, cinematic, photographic, line-art, and a dozen more). Remember the quirk that runs through this whole pack: style_preset is completely ignored unless you flip the style boolean on. It's easy to set a preset, see nothing happen, and assume the node is broken.
• output_format - png (default-safe), webp, or jpeg.

The output is a single IMAGE tensor. It feeds a Save Image / Preview node directly, and since it's a plain image you can chain it into anything downstream - upscale it, img2img it, whatever.

Installing it

This is one node of the "Stability API nodes for ComfyUI" pack. Install via ComfyUI Manager, or:

cd ComfyUI/custom_nodes
git clone https://github.com/Stability-AI/ComfyUI-SAI_API
# restart ComfyUI

No models download with it. The only dependency beyond ComfyUI's own stack is requests, which is already there on essentially every install. You do need a Stability API key - set SAI_API_KEY, drop a sai_platform_key.txt next to the node, or paste into the per-node api_key_override field (which embeds the key in the workflow, so don't share those files).

Common issues

• "payment_required" - no credits left. Trial credits vanish fast; this is a metered service.
• Style preset doing nothing - style is false. That's the pack's design, not a bug.
• Bigger than you expected - this route renders at up to 2016×1152 for 16:9, so if a 4K-looking output surprises you, that's the resolution working as intended.

A no-drama hosted generator. It won't beat a tuned local pipeline for quality-per-dollar, but it's the easiest text-to-image node in the pack to just use.