To use Stable Diffusion, pick how you'll run it: online with no install through Comfy Cloud, through Stability AI's API with pay-as-you-go credits, or free on your own computer with ComfyUI and a downloaded model. Then type a prompt describing the image and generate.
Stable Diffusion isn't one app. It's a family of open image models from Stability AI, and you use it through other software. That's why beginners find it confusing. The good news is that there are only three realistic routes, and you can pick one in a minute. The latest image models are the Stable Diffusion 3.5 family, which Stability lists as Large, Large Turbo and Medium among its Core Models.
Step 1: Choose how you'll run it
| Route | Best for | What you need |
|---|---|---|
| Comfy Cloud | Trying it without installing anything | A Google sign-in; ComfyUI says you get 400 free credits every month |
| Stability AI API | Developers and apps | A Stability account; you start with 25 free credits |
| ComfyUI on your computer | Free, unlimited use on your own hardware | Comfy Desktop and a downloaded model; a graphics card helps |
If you're not technical, start with Comfy Cloud to learn the interface, then decide whether it's worth installing locally.
Step 2: Install ComfyUI if you're running it locally
ComfyUI is the tool Stability AI itself recommends for local use on its Hugging Face model page. Comfy Desktop installs on its own on:
- Windows 10 or later, with an NVIDIA or AMD graphics card recommended.
- macOS on Apple Silicon.
- Linux, as an AppImage or a .deb package.
There's also a portable Windows version that always runs the latest updates, and a manual install that supports more types of hardware.
Step 3: Load a workflow and a model
ComfyUI usually opens the default text-to-image workflow when it launches. Most installs don't include a model, so the first time you run it, ComfyUI warns you that a model is missing, shows which one and gives you the download link. Models live in the ComfyUI/models folder, in subfolders such as checkpoints, lora and vae.
A useful trick: every image ComfyUI makes stores its workflow inside the file. Drag a ComfyUI image into the window, or use Workflows then Open, and you get the exact setup that produced it. That's the fastest way to copy a look you like.
Step 4: Write your prompt
Text-to-image works from two prompts. The positive prompt describes what you want, and the negative prompt lists what you don't. Be specific about the subject, setting, lighting and style.
Copy this prompt
Positive: a bright flat-lay photo of a handmade ceramic coffee mug on a pale oak table, morning sunlight, soft shadows, minimal styling, natural colours
Negative: blurry, text, watermark, extra handles, distorted shapes
Stability says Stable Diffusion 3.5 Large improved typography and understanding of complex prompts, so it handles short words on signs and labels better than older models. Keep any text in the image short.
Models are shared on Hugging Face, which I explain in my guide to what Hugging Face gives you for nothing.
Using the API instead
If you'd rather call Stable Diffusion from your own app or a no-code tool, Stability AI's platform charges in credits. As of October 2026, its pricing page says 1 credit = $0.01 in US dollars, and you start with 25 free credits. Each image costs a set number of credits:
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
| Model on the API | Credits per image |
|---|---|
| Stable Diffusion 3.5 Large | 6.5 |
| Stable Diffusion 3.5 Large Turbo | 4 |
| Stable Diffusion 3.5 Medium | 3.5 |
| Stable Diffusion 3.5 Flash | 2.5 |
| Stable Image Core | 3 |
The same platform offers editing tools such as Remove Background, Inpaint and Erase Object, each priced in credits too.
Customising the look with LoRAs
Once you're comfortable, you'll come across LoRAs: small add-on files that teach a model a particular style, product or character. They go in the lora folder inside ComfyUI's models folder. Stability's licence FAQ confirms that LoRAs, fine-tunes and retrains aren't treated as new foundation models, so you're free to make and use them, and you can even use a LoRA's outputs to train another LoRA. What you can't do is use the Core Models or their outputs to create a new model architecture.
A note on updates: Comfy Desktop follows the stable ComfyUI release by default. If you want every new feature the day it lands, ComfyUI points you to the portable version or a manual install instead.
The licence rule business owners need to know
Running the models yourself is free for most people, but there's a revenue line. Stability's Community License allows research, non-commercial and commercial use for individuals and organisations with under $1M in annual revenue, in US dollars or the local equivalent, regardless of where that revenue comes from. Above that, you need an Enterprise licence. Individual creators making images at home don't pay. Stability also says the Community License can be revoked if you break its terms, so read it before you build a product on it.
For text models on your own machine, see whether Ollama is free.
My verdict
For a small business that wants product mock-ups or social images, start with Comfy Cloud's free monthly credits, then install Comfy Desktop if you find yourself using it every week. Drag in example images to learn from their workflows rather than building from scratch. My Stable Diffusion prompts give you more to try, my guide on how to use ComfyUI goes deeper on the interface, and Midjourney vs Stable Diffusion helps if you're still choosing.
Frequently asked questions
Is Stable Diffusion free to use?
Running the Core Models yourself is free for individuals and businesses under $1M in annual revenue, under Stability AI's Community License. The API is paid in credits.
Can I use Stable Diffusion without installing it?
Yes. Comfy Cloud runs ComfyUI online with a Google sign-in and, according to ComfyUI, 400 free credits every month.
What do I need to run Stable Diffusion on my computer?
Comfy Desktop on Windows 10 or later, macOS on Apple Silicon or Linux, plus a downloaded model. On Windows an NVIDIA or AMD graphics card is recommended.
How much does the Stable Diffusion API cost?
Stability AI says 1 credit = $0.01 as of October 2026. Stable Diffusion 3.5 Large costs 6.5 credits an image and new accounts start with 25 free credits.
Can I use Stable Diffusion images commercially?
Yes, under the Community License if you or your organisation make under $1M a year. Above that, Stability asks you to get an Enterprise licence.
