Every now and then, a new piece of technology creates a radical shift in how people interact with computers. One such shift is currently underway: the rise of AI. Recent breakthroughs have been incredibly exciting to observe. One of the latest innovations is the emergence of generative image models that produce stunning images from text prompts. Because we're obsessed with innovation, we've been eagerly following along here at bunny.net.
At bunny.net, we believe the future brings a fundamental change. Content won't just be cached on the edge. Eventually, it will be generated there too. Today, we're excited to announce our first step in that direction with Bunny AI: an experimental multi-model image generation engine that offers the easiest way to generate images dynamically, directly on the CDN. Generating an image is now as simple as loading a specially crafted URL.

Why Bunny AI?
We've been thrilled by recent AI advancements, and we wanted to give developers the simplest way to empower applications with dynamic image generation. For example, this can be used for profile avatars, game characters or items, or even blog post images, placeholders, and more.
The first and most powerful use case we saw was dynamic profile avatar generation, so naturally, we made bunny.net itself our first real-world use case.
Thanks to Bunny AI, we were able to replace the boring gray profile placeholder with colorful pixel art renders of countless unique bunnies. We use the MD5 hash of your email to generate a random seed for each of our accounts. The result is a cute pixel bunny avatar for anyone not using a Gravatar profile image.

How does it work?
Because we're focused on helping developers build a faster internet, we didn't go out and build our own models. Instead, we partnered with two incredible companies making huge strides in the industry: OpenAI and StabilityAI.
Each offers models with different strengths and weaknesses. We believe the best way to achieve great results is to use the model best suited for the desired output.
DALL-E 2 is fantastic for things like pixel art or more creative prompt renditions, whereas StabilityAI, with its Stable Diffusion model, is great for more photorealistic prompts such as photos or realistic illustrations.
Bunny AI makes it easy to toggle between different "engines" to switch between generative models and resolutions, based on your needs.
Making generating easy with Blueprints
Our main goal was to make dynamic generation as simple as possible and keep URL structures clean. That's why we introduced Blueprints, a feature similar to Image Classes. Blueprints lets you define pre- and post-prompt messages as well as other generation parameters.
For example, we created a "cyberpunk" style that automatically prepends "cyberpunk avatar" and appends "cyber tech with colorful background pastel colors, digital art" to your prompt. You can now generate consistent images of various motives in the same style with a single word like "bunny" or "panda."

This not only creates more pleasing URLs for your end-users, but it also makes them a bit more SEO friendly.
Dynamic processing with Bunny Optimizer integration
To push things even further, Bunny AI Image Generation also integrates with Bunny Optimizer. This allows you to resize, crop, sharpen the images, and more using simple query parameters.
Together, Bunny AI and Bunny Optimizer form a really powerful toolkit for dynamic image generation across many use cases. For example, you can generate a couple of unique images of a bunny pixel avatar and resize them to 128 pixels like this:
https://bunnynet-avatars.b-cdn.net/.ai/img/dalle-256/avatar/email-1/rabbit.jpg?width=128 https://bunnynet-avatars.b-cdn.net/.ai/img/dalle-256/avatar/email-2/rabbit.jpg?width=128 https://bunnynet-avatars.b-cdn.net/.ai/img/dalle-256/avatar/email-3/rabbit.jpg?width=128 https://bunnynet-avatars.b-cdn.net/.ai/img/dalle-256/avatar/email-4/rabbit.jpg?width=128 https://bunnynet-avatars.b-cdn.net/.ai/img/dalle-256/avatar/email-5/rabbit.jpg?width=128





Simple security through Token Authentication
To prevent abuse and unnecessary cost spikes, Bunny AI uses Bunny CDN token authentication. Every image URL must be signed using a specially crafted token. This only takes a few lines of code, and we provide ready-made samples in PHP, C#, JavaScript, and Python.
If you need an example in another language, you can also reach out to our Super Bunnies, and they'll be happy to help.
If you don't want to mess with tokens at all, the simple generator tool inside the panel can pre-sign requests and generates the tokens automatically. And if you have full control over content output, you can also disable tokens entirely using an edge rule.
Try it out yourself!
Bunny AI is currently available free of charge during its experimental preview release and is enabled for every bunny.net user. We invite you to try it out, play around, and share the results with us. Bunny AI is released as an experimental feature, and we'd love to hear your feedback.
To enable Bunny AI, simply toggle the feature in the Bunny AI panel and start generating, or check out the full documentation on Bunny Docs.
Help us make the internet hop faster!
At our core, bunny.net is focused on building a faster internet, and we're working on some exciting projects with deep AI integration. If you're an AI expert who shares our mission and would like to help us shape how people connect to the internet, check out our careers page. We'd love to have you onboard.