Craiyon (formerly known as Dall-E mini) is an online experiment implementing the much more complex Dall-E AI-based image generation from natural text.
Dall-E (sparkled from Elon Musk’s OpenAI) is a limited-access, high-end, AI model that’s already in its second generation and features realistic images and high quality text parsing and matching.
Boris Dayma, a Texas-based machine learning consultant, began from this huge work and developed its own version, at first named Dall-E mini, now turned to Craiyon.
By turning viral, Dall-E mini/Craiyon traffic skyrocketed, according to AI hosting service Hugging Face (more than 50k requests per day).
Crayon AI model relies on some 30 million labeled images and language processing, turning complex phrases into weird-looking images.
You can find Craiyon Python code on Github.
Besides memes and jokes, technology like this bring new arguments to bias, fake generation and how reality is (and will be) perceived).
Twitter, Reddit users crazed about it: I got into this with some early (utterly crazy) attempts (see below).
My advice: the weirdest the text you give Craiyon, the best results you’re likely to get.
Thanks to Andrea for taking me into this machine learning-driven hell 🙂