Accomplice updates: New models, prompt editing, backend upgrades, and a call for creators

Accomplice product updates for March 2022

Welcome to the Accomplice product updates for March 2022. There's a lot to cover so ready to just jump right into it? 👇

New default model: “Disco Diffusion”

“floating dystopian steampunk city” and “rain falling in the alleyway of a dystopian city at night lit by neon lights” created using “Disco Diffusion”

Default models are the ones every new Accomplice user sees when they sign up and create their first image. They’re the workhorses that can do a little bit of everything. ImageNet is the GOAT. WikiArt is a close second. Style Transfer can do things no other model can. And, when you really just need an actual goat – like you know the animal itself – GLIDE is there for you.

As of today, Disco Diffusion has been added to that default list of models, and I think it will quickly become your new favorite. It takes modifiers – like “by Claude Monet” and “by Megan Duncanson” – just like ImageNET and WikiArt do. It takes image prompts. And it creates large, truly unique images that you won't be able to get enough of.

I mean just look at some of these examples created using Disco Diffusion and then go try it out for yourself!

city on fire at night
ocean waves crashing on beach rocks
pink japanese cherry tree watercolor

4 more amazing new models

Accomplice already has the largest selection of image models in the world at over a dozen and counting. Over the coming months that number will continue to grow each and every week. My goal for later this year is that everytime you use Accomplice there will something new for you to create with – starting with these 4 new models below!

Diffusion VI

Diffusion VI

Diffusion VI is the little brother to Disco Diffusion. It’s not quite as consistently good as big brother, but it creates 6 images at a time, and oftentimes creates truly one-of-a-kind images. It's a great addition to your Accomplice toolbelt. It can take the style, the content, or both from an image prompt, and does an amazing job at colorful, abstract patterns.

Colorful Blobs AI

Colorful Blobs AI

The Colorful Blobs AI model is the first model created by an Accomplice user! @vinsent is offering it for free and it's a fun, colorful, abstract blob model great for backgrounds and presentations.

(If you’re interested in creating models, either for free or to sell, jump down to the end of this post for more information or just reach out to

Product Spot AI

Product Spot AI

I'll be honest, Product Spot AI is… a work in progress. I loaded over 1,000 open source illustrations into Accomplice to see how good it could get at creating landing page illustrations. And while it's a little hit or miss, it's still a robot that makes you vector-ish illustrations of whatever you tell it to, so it's fun and worth a try regardless – and I'll keep working on it to see just how good it can get at it!

Butterflies AI

While the dream of one huge, perfectly versatile image model that can do whatever you tell it is tempting – I believe it is just that. A dream.

Instead of one, giant, "super model", I think specialized image models like this one – Butterflies AI – are the future. Not only of Accomplice, but of design in general.

Have a project where you need flawless, detailed, beautiful butterflies? Instead of buying one stock photo or illustration for $10, spend $10 worth of Accomplice credits using the Butterflies AI model to make a butterfly that is any style and combination of colors you need simply by typing what you want and hitting create. And, best of all, you can use the model over and over again. So next time you need it, you can use it to knock out another butterfly… or 20.

You can create a model that makes endless results as good as the above for almost anything you can think of. The possibilities really are endless.

(And if you think you have an idea for a model you'd like to create, jump to the bottom of the post or reach out to

New feature: Prompt editing

The new Edit button

With any of the models that create one image at a time – ImageNet, WikiArt, Style Transfer, and now Disco Diffusion – only the latest iteration is ever shown at a time. But now, if you want to get to one of the earlier iterations, you can select the new "Edit" button at the top of the prompt screen, and the sidebar will pop out showing you all the iterations that were created during the building of that image.

The last iteration created is always "kept", and you can choose to delete or keep any of the other iterations as well. Any iterations that you don't delete or keep will get automatically deleted 30 days after the creation of the image.

Also, now, when editing a prompt you can make any iteration you want the prompt’s "thumbnail", and that is the image you will see when you browse through your prompts in the Inbox or Collections.

Backend upgrades: Accomplice is now faster and creates higher quality images

colorful arrows going up watercolor

Over the weekend, Accomplice got a major upgrade to its backend. Everything is now not only a little bit faster, but also higher quality. The images that the ImageNet and WikiArt models create are now about 25% larger than they were before, so they should look better than ever. All the Accomplice-powered models are faster and higher quality. And Style Transfer and GLIDE are now done almost before they even begin.

(And stay tuned because more and even better upgrades like this are coming soon!)

In addition to the overall upgrade, paid subscriptions now have the option of a "High quality" setting. This means at least 200 more iterations for each model that offers it. The highest quality costs an extra credit, but the results are really great.

Help wanted: Create your own public or premium model for free

I mentioned above that, at just over a dozen, Accomplice has the largest selection of image models in the world. But I need your help to create even more!

Accomplice already makes it easy to build your own custom models. And now I want to open up that feature to anyone who has an idea for a free or premium Accomplice model that other Accomplice users will want to create with.

If you have an idea for something that would make an awesome Accomplice-powered model – like Butterflies AI or Low Poly AI or Avatars AI – reach out to