You might be trying to dip your toes into AI image generation with Stable Diffusion.
Installing it onto your computer or running in the cloud with Google Colab notebooks sound great and all, but you're wondering if there's anything that lets you...just get started?
Here's a number of websites I've tested that'll let you get started prompting instantly.
They'll be divided into 3 sections:
- Websites that are completely or mostly free
- Websites that are paid, but have some free features
- Websites that are paid-use only
TDLR:
- Of all the websites I've tested, my favorite is PlaygroundAI. The free plan gives you 1000 generations a day, it has tons of custom models and a good onboarding/teaching process. It's by far the best way for beginners to get started.
- If you want more extensive Stable Diffusion features (such as ControlNet) for free, use ArtBot or Stable UI. Prepare for long generation times.
- Amongst paid websites, I prefer Mage.space for its social features, and Getimg.ai for its photo editing features (which are similar to Photoshop's generative AI features)
- RunDiffusion is my preferred cloud service for running AUTOMATIC1111, if you want the highest levels of customizability and the latest and greatest features. I would only recommend this if you've used AUTOMATIC1111 before.
Completely/Mostly Free Websites
PlaygroundAI
This is one of the more famous apps. They've raised quite a bit of money, and have much fancier marketing/slicker branding than the rest of the apps on this list, so I wanted to see what the fuss was all about.
- Features: A
- Ease of use: A
- Generation Speed: A
- Pricing: Free up to 1000 images a day, $12/mo annually for 2000 images a day + advanced features
- Sign-up: Required
- NSFW: Censored
- Comes with: Infinite canvas, img2img, wide model selection, community feed
Woah a triple-A score? I was surprised too.
PlaygroundAI is free for up to 1000 images a day, but most people aren't going to get anywhere near that. So this is my number one recommendation for people who are trying to get started with Stable Diffusion.
The onboarding process is well thought out. Since new users aren't going to familiar with Stable Diffusion concepts and vocabulary, Playground rolls them out gradually and uses simplified terms. Some choices are interesting: models are called "filters", which are easier to understand from the end user perspective.
The biggest difference between PlaygroundAI and the other websites is the Canvas view. You can pan or zoom around a canvas that you put multiple generations on: this feels like something straight out of Photoshop or Figma.
There are very few options as to what you can actually do with these images (remove background, img2img), but it dawned on me that this is the only the beginning.
PlaygroundAI is trying to create the Photoshop of the future, so I have no doubt many more features will be added over time.
ArtBot
- Features: A
- Ease of use: B
- Generation Speed: D
- Pricing: Completely Free
- Sign-up: Not required, but you can for the social features
- NSFW: Uncensored
- Comes with: img2img, wide model selection, social gallery, ControlNet, inpainting
ArtBot is a Stable Diffusion website based on the Stable Horde. The Stable Horde is a crowdsourced cluster of GPUs. Anybody can be generous contribute their GPU to the horde so that other people can use them.
This is a very well done project by Dave Schumaker.
There is a kudos system: by voting on other peoples' generations or contributing your GPU to the Stable Horde, you gain kudos, which will give you priority in the generation queue.
If you don't have kudos, generations take quite a while: it took me 7 minutes to run a generation.
Stable UI
- Features: A
- Ease of use: B
- Generation Speed: D
- Pricing: Completely Free
- Sign-up: Not required
- NSFW: Uncensored
- Comes with: img2img, inpainting, wide model selection, interrogation (upload an image and have it described)
Another Stable Horde based website. Also has a kudos system.
My queue time was 6 minutes to run a generation.
Paid Websites with Free Features
Getimg.ai
- Features: A
- Ease of use: A
- Generation Speed: A
- Pricing: 100 free images a month, $12/mo for basic plan 3000 free images a month
- Sign-up: Required
- NSFW: Uncensored
- Comes with: Img2img, controlnet, outpainting, image editing
In terms of taking Stable Diffusion features that might be a little bit complicated, and making them easy to understand for most people, Getimg.ai is leading the pack.
Getimg.ai has some unique features that differentiate it from other websites:
AI Canvas: This is an infinite canvas that lets you use outpainting as much as you want. Outpainting is when you generate more of an image, outside of its frame.
Image Editor: This is close to Photoshop's generative AI functionality. The image editor lets you inpaint (paint over an area and regenerate it), as well as run other Stable Diffusion features on the image.
DreamBooth: DreamBooth is the technology behind the AI Avatar trend. It's what lets you upload your image and generate images of yourself in any style. Few of the websites on this list have this feature. It's only available on paid plans.
DreamStudio
This is the "official app" by Stability AI, the creators of Stable Diffusion.
- Features: C
- Ease of use: A
- Generation Speed: A
- Pricing: $10 per 1,000 credits (~500 generations)
- Sign-up: Required
DreamStudio is easy to use, it does a good job of surfacing core Stable Diffusion features (text-to-image) and (image-to-image), and gives you 25 free credits, which is roughly 12 images. These credits are used interchangeably with the StabilityAI API.
You would think that since StabilityAI is the creator of Stable Diffusion, DreamStudio would be a pretty sophisticated app, but it isn't. It appears Stability AI isn't focused on consumer apps (hence their acquisition of Clipdrop.co, another Stable Diffusion app featured in the next section) because they're too busy making models.
One cool feature is the style presets. These add words to your prompts to create specific styles.
The biggest drawback is you're limited to the base models (SD v1.x, SD v2.x or SDXL). There's no way to upload or use any custom models.
Clipdrop.co
Clipdrop is StabilityAI's other image generation app. The difference between this one and DreamStudio is that this one is an entire suite of tools: most are powered by Stable Diffusion, the rest are fairly standard image-editing tools.
- Features: C (Clipdrop actually has many features, but I'm talking about features within Stable Diffusion)
- Ease of use: A
- Generation Speed: A
- Pricing: $5/month
- Sign-up: Required
When SDXL was first announced, Clipdrop was the only way to access the beta. We're assuming StabilityAI will use Clipdrop to showcase whatever they build in the feature.
The downside is the actual Stable Diffusion generation does not come with many features.
Some of the Stable Diffusion-powered tools you might be interested in:
Reimagine XL: Create variations of the same photo. If you use a photo of a person, you'll get different angles and compositions of the same person.
Stable Doodle: Upload or draw something, then create generations from it.
Mage.Space
- Features: B
- Ease of use: A
- Generation Speed: A
- Pricing: Free, $15/mo for Pro plan which comes with faster GPUs and advanced features
- Sign-up: Not required but encouraged
- NSFW: Uncensored
- Comes with: Social gallery, img2img
A bit part of this website is the gallery: you can see other peoples' generations, rerun them or remix them with an altered prompt or settings.
Dezgo
- Features: A
- Ease of use: B
- Generation Speed: A
- Pricing: Free but with severe restrictions, uses a credit-system: with $10 you can generate roughly 558 SDXL images
- Sign-up: Not required but encouraged
- NSFW: Uncensored
- Comes with: Controlnet, img2img, upscaling, SDXL, text-to-video (AnimateDiff), wide model selection (thought not the most up-to-date)
You pretty much have to pay for Dezgo: the limitations are too severe otherwise. That said, it's a decent app if you have credits.
Dreamlike.art
- Features: C
- Ease of use: A
- Generation Speed: A
- Pricing: 50 free credits, $12/mo plan with 3000 credits. Generations cost ~2.2 credits.
- Sign-up: Required
- NSFW: Uncensored
- Comes with: Generation history, img2img, wide model selection, face restore, upscaling
Dreamlike.art has an easy to use interface, but is not as full featured and cost efficient as the other websites on this list.
I suspect you hear about them a lot because they have an affiliate program.
Paid Websites
RunDiffusion
- Features: A
- Ease of use: C
- Generation Speed: A
- Pricing: $0.50/hour
- Sign-up: Required
- NSFW: Uncensored
- Comes with: Everything
This is quite different from all the other websites, because it runs AUTOMATIC1111 on a virtual machine. So instead of purchasing credits, you're purchasing hours of usage. In that sense, it's more similar to Google Colab and other cloud GPU providers, with slightly more handholding.
I would only recommend this if you're an intermediate Stable Diffusion user who has used AUTOMATIC1111 before.
The benefit is that since you are running AUTOMATIC1111, you have access to the latest and greatest features, extensions, and things like video (which the other websites on this list have not implemented, or only implemented a very simple version of).
Wrapping Up
There's a number of apps to choose from, and your selection will ultimately depend on your needs.
If you're just trying to generate images from text and use multiple styles, PlaygroundAI is the best bet.
As you go from a beginner to intermediate Stable Diffusion user, you'll want to use more advanced features. In this case, I recommend you try a WebUI such as AUTOMATIC1111, which you can install locally or use with a cloud GPU service such as Google Colab.
After you decide what website to use, you attention will probably turn towards prompting. Check out the prompt guide and settings guide. The prompt guide will be relevant regardless of which website you use, the settings guide may or may not be relevant, since many of these websites hide these settings from the user, or use other names for them to make them easier to understand.
