Stable Diffusion
Stable Diffusion 3.5 is a family of open text-to-image models from Stability AI that you can download, customize and run on consumer GPUs.
Key Takeaways
Open image models you can download, fine-tune and run yourself
Free · $0
- Developers building image apps
- AI hobbyists and tinkerers
- Researchers
What is Stable Diffusion?
Stable Diffusion is Stability AI’s family of text-to-image models. The 3.5 release comes in three versions: Large, with 8.1 billion parameters for top quality at 1 megapixel; Large Turbo, a distilled version that produces images in just 4 steps; and Medium, at 2.5 billion parameters, designed to run on everyday GPUs.
The weights are available on Hugging Face and the inference code on GitHub. The models were built to be fine-tuned, so you can train LoRAs, build custom workflows or ship your own apps. They handle styles from photography and 3D to painting and line art.
They are released under the Stability AI Community License, which allows free commercial and non-commercial use for most users. Researchers, hobbyists, startups and enterprises are all named as audiences.
Who is Using Stable Diffusion?
Developers and startups
Build image features and apps on open weights.
Hobbyists
Run image generation locally and experiment with LoRAs.
Researchers
Study and extend open image models.
Key Features
- Stable Diffusion 3.5 Large: An 8.1 billion parameter model with strong prompt adherence at 1 megapixel.
- Stable Diffusion 3.5 Large Turbo: A distilled model that generates images in just 4 steps.
- Stable Diffusion 3.5 Medium: A 2.5 billion parameter model for consumer GPUs, producing 0.25 to 2 megapixel images.
- Customizability: Query-Key Normalization helps make fine-tuning and further development more stable.
- Diverse outputs: Produces a range of people and styles without heavy prompting.
- Versatile styles: Covers 3D, photography, painting, line art and many other looks.
Pricing Plans
Community License
- Free for most commercial and non-commercial use
- Stable Diffusion 3.5 Large
- Stable Diffusion 3.5 Large Turbo
- Stable Diffusion 3.5 Medium
- Weights on Hugging Face and code on GitHub
Which Stable Diffusion plan should you pick?
Developers, researchers and creators running models themselves.
Stable Diffusion Use Cases
Local image generation
Run Medium on a consumer GPU.
Outcome: Image creation without sending prompts to a hosted service.
Fine-tuning a style
Train a LoRA on your own images.
Outcome: A model that matches a brand or art style.
Fast drafts
Use Large Turbo for quick iterations.
Outcome: Many candidate images in a short time.
How Stable Diffusion fits into your stack
Hugging Face and GitHub
Model weights are on Hugging Face and inference code on GitHub.
Amazon SageMaker
Stable Diffusion 3.5 Large is available in Amazon SageMaker JumpStart.
Stability AI deployment
Stability AI also lists a Platform API and a Self-Hosted License.
Pros
- +Free under a permissive license: Commercial and non-commercial use is allowed under the Community License.
- +Runs on consumer hardware: Medium needs about 9.9 GB of VRAM, excluding text encoders.
- +Built to customize: The models are designed for fine-tuning and LoRA training.
- +Choice of size: Large, Large Turbo and Medium trade quality against speed.
Cons
- −Needs technical setup: You download weights and run inference code yourself.
- −Output can vary: Vague prompts may give less predictable results and mixed aesthetics.
Limitations & when to consider alternatives
- No hosted app on this page. The source covers model downloads, not a ready-to-use web editor.
- License terms apply. Use is governed by the Stability AI Community License, so larger organizations should check the terms.
- You want a polished web app with no installation
- You do not have a GPU and prefer a simple subscription
How to Use Stable Diffusion
Pick a model
Choose Large, Large Turbo or Medium based on quality and hardware.
Download
Get the weights from Hugging Face and the inference code from GitHub.
Run it
Load the model on your GPU and start generating from prompts.
Customize
Fine-tune or train LoRAs for your own needs.
Who is Stable Diffusion For?
Technical users
People comfortable installing and running AI models.
Product builders
Teams that want to embed image generation in their own products.
Frequently Asked Questions
Is Stable Diffusion free?
The 3.5 models are free for commercial and non-commercial use under the Stability AI Community License, which covers most users.
What hardware do I need?
Medium needs about 9.9 GB of VRAM, excluding text encoders, which fits most consumer GPUs.
Where can I download it?
The weights are on Hugging Face and the inference code is on GitHub.
Which version should I use?
Large gives the best quality, Large Turbo is fastest, and Medium balances quality with easy setup on consumer hardware.
Who makes Stable Diffusion?
Stability AI, which also makes Stable Audio.
Stable Diffusion Alternatives
Compare allLooking for alternatives to Stable Diffusion? Here are similar tools in the same category:
Pixelate by Scenario
Turn any image into pixel art with custom grids and palettes
P
ProfilePicture.AI
Make a clean profile picture in your browser for free
Palette
Add realistic color to black and white photos in seconds
Compare Stable Diffusion with: Stockimg AI Scribble Diffusion REimagine Home Pixelate by Scenario