Two weeks of Stable Diffusion, monthly credit refresh, pricing updates and RIP old models

Labor Day edition of the latest and greatest Accomplice product updates

Two weeks of Stable Diffusion

It's not even been 2 full weeks since Stable Diffusion came out of Beta. It was just last Monday! And I announced Accomplice support for it later that same day.

Then, a few days later on Twitter, I announced init image and sketch-to-image support – which caught a lot of people’s attention!

Since that Monday launch, Accomplice users have created over 10,000 images with Stable Diffusion.

Go browse all the amazing shared Stable Diffusion images on the Accomplice Community by visiting this link.

Plans and credits updates 💳

If you have credits in your account, you may have noticed they quadrupled this week. Or you may have noticed the credits for each plan jumped way up.

To create more flexibility in the coming future – when new editing features like inpainting and outpainting get released – each image now costs 1 credit.

If you create a prompt with 4 images, that now costs 4 credits (and very soon you'll be able to create as few as 1 image per prompt with Stable Diffusion, DALL-E Mini, and all the custom models).

So, since previously each prompt was just 1 credit no matter how many images it created, the credits in your account from any previous plans or purchases have been 4x'd.

(Last note: since the older diffusion models like Disco Diffusion, Pixel Art Diffusion, etc. take more time to create images and create several iterations of the image as it's getting created, they cost 4 credits as well.)

RIP Imagenet and Wikiart 😢

Imagenet was the very first model that Accomplice launched with back in September 2021. And for all the people reading this who remember that original launch a) can you believe it's almost been a whole year?! and b) thanks so much for all your support over these 11 months and change.

But a year is an eternity in the AI image space. And not only were Imagenet and Wikiart starting to hold back some of the things Accomplice needed to do on the backend – they were disappointing new users who were expecting higher fidelity models like Stable Diffusion and DALL-E Mini.

So, as a result, the Imagenet and Wikiart models have been archived – along with the old GLIDE preview release that has since been well surpassed by DALL-E 2.

All your old images will always stick around, but no new images can be created with those models.


– Adam