Changes

In 2021, after the release of OpenAI’s [[CLIP]] (Contrastive Language-Image Pre-training), designed to assess the alignment between text descriptions and generated images, there was an increase in popularity for creating high-quality images from [[AI models ]] based on text input. This led to a period of exploration and development by the art community, sharing [[Python ]] code and using various models and techniques to create images <ref name=”1”></ref>.

Midjourney’s technology is different from other image creation tools like [[Adobe Photoshop ]] or [[Blender]]. It exists as a [[social app]], using as its front-end Discord. Users log in to Midjourney’s server, submit text prompts, and an image is generated and shared within the server after, more or less, a minute of computation. Due to its social nature and its creator’s goal of it being a social space for anyone over the age of 13, it is subjected to rules about allowable content, with some text inputs being blocked automatically. The company has around 40 moderators monitoring on the user’s images generated content. For example, the creation of images with adult content or gore isn’t permitted. Also, it is advised that user’s refrain from making overall disturbing or shocking content <ref name=”1”></ref>.

The program output, like DALL-E, results from AI models training on a large number of images. The third version of Midjourney has been introduced with enhancements in image quality coming from it’s social aspect, incorporating “a feedback loop based on user activity and response.” According to [[David Holz]], founder of Midjourney, “It's mind-bogglingly better and we didn't actually put any more art into it. We just took the data about what images the users liked, and how they were using it. And that actually made it better" <ref name=”1”></ref>.