Stable Diffusion in the Cloud: No GPU Required

Who This Is For

You don't have an NVIDIA GPU. Maybe you're on a laptop, an older desktop, an Intel Mac, or a Chromebook. That's fine — you can still generate images using cloud services that give you a GPU over the internet.

The tradeoff: you don't own the hardware, so you're either paying per hour or using limited free tiers. But it works, and the quality is the same as running locally.

Option 1: Google Colab (Free Tier Available)

Google Colab gives you a free GPU in your browser. It's the fastest way to try Stable Diffusion without installing anything.

How to set it up:

1. Go to Google Colab
2. Search for "Automatic1111 Colab" on GitHub or Reddit — there are several maintained notebooks. Look for ones with recent updates and lots of stars.
3. Open the notebook in Colab
4. Go to Runtime → Change runtime type → GPU
5. Run each cell in order. The notebook will install everything, download a checkpoint, and launch the UI
6. When it's ready, you'll get a Gradio link — click it to open the UI

Pros:

• Free (with limits)
• No installation on your computer
• Works from any device with a browser

Cons:

• Google limits free GPU time — you might get disconnected after 30-60 minutes of inactivity
• Free tier GPUs are slower than paid
• Google has been cracking down on SD usage on free Colab. If it stops working, you may need Colab Pro ($10/mo) or another option below

Heads up: Free Colab availability changes. Google sometimes restricts what you can run. If the notebook won't start or you keep getting disconnected, it's not you — it's Google tightening the free tier. Try one of the other options.

Option 2: RunPod (Pay Per Hour)

RunPod rents you a cloud GPU by the hour. More reliable than free Colab and much faster.

How to set it up:

1. Create a RunPod account
2. Go to Templates and search for "Automatic1111" or "Stable Diffusion WebUI"
3. Pick a GPU — an RTX 3090 or RTX 4090 is ideal. Costs around $0.40-0.75/hour.
4. Deploy the pod
5. Once it's running, click Connect and open the web UI

Pros:

• Fast — cloud 3090/4090 GPUs are as fast as local
• Reliable — no random disconnections
• Pre-built templates mean zero setup

Cons:

• Costs money (but $0.40/hour is cheap for experimentation)
• Your files aren't saved when you stop the pod (unless you attach persistent storage)

Tip: If you're just starting out, spin up a pod for 2-3 hours, generate a bunch of images, download the ones you like, then stop the pod. A session costs $1-2.

Option 3: Paperspace (Pay Per Hour / Free Tier)

Paperspace is similar to RunPod but has a free tier with limited GPU access.

How to set it up:

1. Create a Paperspace account
2. Create a new Notebook (their version of a Colab-style environment)
3. Select a Free GPU or a paid one
4. Use a Stable Diffusion template or install A1111 manually in the terminal
5. Launch the web UI

Pros:

• Has a free tier (limited)
• Persistent storage — your files stay between sessions
• Good for longer experiments

Cons:

• Free GPUs have a queue — you might wait
• Interface is less polished than RunPod for SD specifically

Option 4: ThinkDiffusion (Easiest, Subscription)

ThinkDiffusion is Automatic1111 pre-installed in the cloud. You just log in and start generating. No setup at all.

Pros:

• Zero setup — A1111 is already installed and configured
• Extensions and models pre-loaded
• Built specifically for Stable Diffusion users

Cons:

• Subscription based (starts around $8/mo)
• Less control over the environment

Which One Should You Pick?

If you're serious about this and plan to generate regularly, I'd recommend either RunPod (most flexible) or saving up for a local GPU. Cloud costs add up over time, and a local setup is faster and has no limits.

Once You're In the UI

No matter which cloud option you chose, the UI is the same Automatic1111 interface. Everything from here works the same as the Windows or Mac guides:

1. Select a checkpoint from the dropdown
2. Type a prompt
3. Set steps to 20-30
4. Set the size — 512x768 for portrait (SD 1.5) or 832x1216 (SDXL)
5. Click Generate

If your cloud template didn't come with a checkpoint, download one from Civitai:

For anime: Anything V5 or MeinaMix For realistic: Realistic Vision or epiCRealism For stylized: DreamShaper or RevAnimated

Upload the .safetensors file to the models/Stable-diffusion/ folder in your cloud environment.

What's Next?

Once you're generating:

• Why I don't start with a prompt — the mindset shift that changes everything
• The Three Starting Points — how to go from reference images to better prompts
• How I Remix Any Prompt — how to take someone else's prompt and make it your own

Welcome to the rabbit hole.