Stability SD3 Image Generation

Here's a node that can't avoid its own backstory. StabilityImageSD3 generates images with Stability's hosted SD3 models - sd3-medium up through sd3.5-large and sd3.5-large-turbo - via /v2beta/stable-image/generate/sd3. It's the one node in this pack where you could plausibly run the same model locally instead, because SD3.5 Large is open weights. But "could" is doing a lot of work: an 8.1B MMDiT with a triple text encoder is a real chunk of RAM, and this node sidesteps all of it for the price of credits.

The backstory: SD3 Medium's June 2024 launch is the release that ended Stability's run as the open-source leader - broken anatomy out of the box, a revocable license, and an ecosystem that never formed. The API models you're actually using here (SD3.5, released October 2024) are competent, especially large and large-turbo: strong composition and color, decent photorealism for a hosted model. They're just not anyone's default, and the license's acceptable-use policy has flatly banned sexual content since mid-2025. If NSFW work matters to you, this API - and honestly this whole pack - isn't the platform.

How it works

Pick a model and a mode, and the inputs reshuffle themselves around the mode:

• text-to-image - prompt in, image out; aspect_ratio (seven ratios, 1:1 default) controls the canvas.
• image-to-image - you must also wire an image in, and strength (0–1, default 0.7) sets how much of the original survives. Forget the image in this mode and the node errors before spending anything.

Shared: cfg_scale (default 7 - classic classifier-free-guidance feel), negative_prompt, seed, style_preset (a shorter 10-item list here: enhance, anime, photographic, digital-art, comic-book, pixel-art, cinematic, 3d-model, origami), output_format (png/webp), and api_key.

Output: a single IMAGE.

Installing it

Part of the aicu-comfyui-stability-ai-api pack:

cd ComfyUI/custom_nodes
git clone https://github.com/aicuai/aicu-comfyui-stability-ai-api.git
cd aicu-comfyui-stability-ai-api
pip install -r requirements.txt

Or ComfyUI Manager → search "aicu-comfyui-stability-ai-api" → install → restart. Paid Stability AI key required: node api_key field, STABILITY_API_KEY env var, or the pack's auto-generated config.ini.

When you'd actually pick it

Be honest with yourself: if you just want a hosted image, Core is cheaper and Ultra is better-looking - this node sits between them for a reason, and that reason is "you specifically want the SD3.5 look." The strongest use case is comparison work: you're running SD3.5 locally and want a clean reference output from the same family, or you're deciding whether the open 3.5-large is worth the VRAM before committing. The turbo models (sd3.5-large-turbo, sd3-large-turbo) are the distilled fast paths if you want it snappier and cheaper. The model family survived its launch; the ecosystem never did, and that's the honest context for what this node offers.