Boris Dayma designed an open alternative to DALL-E by the name of DALL-E Mini. As the name suggests, it is a “mini” version of the former. The DALL-E Mini is an image generator that employs artificial intelligence to generate a series of images depending on user-entered prompts. It can visualize most weird notions or events in a photo grid format.
According to programmers and researchers, large language-based datasets can train a vast neural network to perform diverse text production tasks and generate high-resolution images. In January 2021, OpenAI developed DALL-E, a multimodal generative neural network to create images from text.
DALL-E uses the trained GPT-3 parameter version to create images using text-image pairs. It is the world’s first creative artificial intelligence model that can generate visuals from the text. In other words, it blends natural language understanding with the generation of realistic pictures.
Following its success and mettle, OpenAI released DALL-E 2, an updated version in April 2022. This version incorporated higher resolution and lower latency. However, DALL-E is not open to the public. Thus, the DALL-E Mini is based on the same concept devised by OpenAI but is free and open to the public. Unfortunately, there is not much that DALL-E mini offers against DALL-E.
Both DALL-E and DALL-E Mini share the same architecture. It is only during the training that the mini version enters a much larger dataset. DALL-E 2 works on a 3.5 billion parameter model and another 1.5 billion parameters to improve the resolution of produced images.
Boris Dayma, the creator of DALL-E Mini, said, “The model is trained by looking at millions of images from the internet with their associated captions. Over time, it learns how to draw an image from a text prompt. Some of the concepts are learned from memory as it may have seen similar images. However, it can also learn how to create unique images that don’t exist such as “the Eiffel tower is landing on the moon” by combining multiple concepts together.”