Accomplice product updates for February 2022
All the new models and features I'm about to go over in this post – as well as tons of tips and tricks and things you might not have known about from past releases – are also available in the new Accomplice Quick Start Guide.
At the end of December 2021, OpenAI unveiled GLIDE, but only in a "preview release". (As a result it doesn't yet support people and the images start off really small.)
But, even despite its preview release limitations, its animals and photo-realistic places are already better than anything that has come before it!
Also? GLIDE creates 6 images at a time. And it's fast. So if you're looking for lifelike animals or places, give GLIDE a try.
Note: GLIDE doesn't respond to modifiers like "by Van Gogh"
In addition to being more capable at producing images, Accomplice is also going to start becoming more capable with its post-production editing capabilities. The first step: the new Style Transfer model.
With Style Transfer you can take the "style" of an uploaded image – so its repeating background, or its colors and stripes, or its brush strokes, etc – and transfer it to the "content" of another uploaded image – like a photo of a person or place, or a logo, or anything really.
Several creators are already creating both the style and the content images in Accomplice, downloading them, and then reuploading them to use them with the Style Transfer model!
Models like ImageNET, WikiNET, and GLIDE are extremely versatile, and that's useful when you want something that can do a little bit of everything.
But what if you also had models that only did one thing, but amazingly well?
Enter Accomplice’s new, exclusive, Accomplice-powered models.
There are currently 6 new Accomplice-powered models that you can activate and start using today. And, starting this week, there will be new ones released every week or so.
Useful? Not necessarily. Super fun? Absolutely!
The Pokemon AI model was trained on all 900+ Pokemon in existence. So, if you can think it, most likely Accomplice can create you a Pokemon for it.
Low Poly AI
From the not necessarily useful Pokemon, to the extremely useful low Low Poly AI model.
Headers. Backgrounds. Websites. Apps. Crisp, colorful triangles you can use practically anywhere for practically anything.
Originally designed by Pablo Stanley, Avataaars are a fun, diverse set of vector avatars you can mix and match.
Great for social media, mockups, ads, and the like.
Skulls AI, Waves AI, and Emoji AI
Rounding out the new Accomplice-powered models are Skulls AI, Waves AI, and Emoji AI.
Skulls AI was trained on 50 different styles of black and white, with dashes of color, hand drawn skulls.
Waves AI was trained on 100 different styles of colorful wavy backgrounds.
Emoji AI was trained on all 2000+ emojis currently in existence. It can do just about anything in the style of an emoji.
The really awesome news about all the above Accomplice-powered models? They were trained using the same drag and drop interface that Pro subscribers and above also have access to!
Want to train Accomplice to create more marketing materials for your company based on all your old marketing materials? Working on a project that requires a bunch of something all in the same unique style?
Create a model, add a bunch of examples images in the style you want Accomplice to create more of, select "Build model", and in several minutes – and after a short review process – you will have your very own image model that will do more of what you taught it over and over again!
There will be more information and resources about the power of custom models coming very soon. But in the meantime play around with it and see what you can create!
Note: If you are a large company or corporation and would like help building your own custom models for your team please reach out to firstname.lastname@example.org
Don't yet have the need for custom models? Or is it just you and you don't need teammates yet? Or only need to create images here and there throughout the month? Then the new Personal plan is for you!
Whew. If you got this far and are reading this – I appreciate you. 👊🏼