What is DALL-E?
The DALL·E 2 algorithm is an evolution of the original DALL-E by OpenAI. DALL-E kicked off the whole text-to-image scene by demonstrating that it was possible. DALL-E 2 brings a whole new level of coherence, resolution, and prompt accuracy.
DALL-E is an AI (Artificial Intelligence) system that has been designed and trained to generate new images. The technology can generate an image from a text prompt, like "An astronaut lounging in a tropical resort in space in a photorealistic style" (below).
Prompt: "An astronaut lounging in a tropical resort in space in a photorealistic style". Image credit: OpenAI
While other text-to-image systems exist (e.g. VQGAN+CLIP and CLIP-Guided Diffusion, which are tokens-based programs that are available on NightCafe), the latest version of DALL-E is much better at generating coherent images. The technology seems to have a good understanding of the world, and the relationships between objects. DALL E can create beautiful images that show what AI is capable of creating.
Is DALL-E Available to the Public?
Yes! DALL-E 1 was first created by OpenAI in early 2021, and the model was never released to the public. In April 2022, OpenAI announced DALL-E 2, which started in private beta status but was eventually released to the public.
While DALL-E was still in a closed private beta status, other open-source DALL E equivalents like Stable Diffusion were also released to the public. Both DALL E and Stable Diffusion are available on NightCafe.
Who Owns DALL-E?
DALL-E was invented by OpenAI, a company dedicated to AI research and ensuring that AI technology benefits all of humanity.
Who Owns OpenAI?
OpenAI consists of both a for-profit and not-for-profit company. It was founded in 2015 by Elon Musk, Sam Altman, and others who pledged $1 Billion. Current investors include Microsoft, Reid Hoffman, Khosla Ventures, and others. Elon Musk resigned from the OpenAI board in 2018.
When Can I Generate Art with DALL-E Online?
As of October 28, 2022, you can use NightCafe to generate images with DALL-E 2 online. You can start generating images for free and your credits will replenish daily or you can purchase more credits to continue creating!
Is DALL-E Real?
It's hard to believe that all the images on this page were generated by AI from nothing but the text prompts shown, but it's true! The images on this page are an example of what is possible with DALL E AI. DALL-E is indeed real and is an exciting glimpse into the future of creative AI.
How Does DALL-E 2 Work?
DALL-E 2 uses a similar (but more advanced) method to NightCafe's "Coherent" image creation algorithm. DALL E 2 works by starting with an image made up of randomized pixels ("noise") and gradually "de-noising" the image. Throughout the de-noising process, it is guided towards an image that matches the initial source prompt.
Every DALL-E AI generated image generated on NightCafe will be unique and you are effectively its "author". You're free to use the images for any purpose you choose.
How Was DALL-E 2 Trained?
According to OpenAI, DALL-E 2 was trained on up to 650 million text-image pairs from a variety of sources. Explicit content (e.g., porn, gore, etc) was filtered out of the training data to prevent DALL E from being good at producing those kinds of pictures.
What is GPT-3?
GPT-3 is another AI system developed by OpenAI, which is good at Natural Language Processing (NLP). Unlike the GPT-3 text generator, DALL-E was trained by, and is used to create, images rather than text.
GPT-3 is different from DALL E because it generates text rather than images. GPT-3 can generate text that is very difficult to distinguish from human-written prose. GPT-3 has been used to generate poetry, news articles, short stories, and even software code (Github Copilot).
Is GPT-3 a Generative Adversarial Network (GAN)?
No. GPT-3 is a generative model, but it does not use an adversarial network. Instead, GPT-3 is based on a neural network architecture invented in 2017 called a Transformer.
What can I use DALL-E for?
You can use DALL-E to generate images of anything you can describe in text. For example, you could use DALL-E to generate an image of a robot, a flower, or a car.
DALL-E can be used to generate images from textual descriptions, similar to how a human might draw an image from textual descriptions.
Some potential applications for DALL-E include generating product images for prototyping, designing new buildings and structures, generating art, creating new ways to visualize data, and much more.
Is DALL-E 2 The Best Text-to-Image AI?
According to many model benchmarks, DALL-E 2 achieved state-of-the-art results in text-to-image generation when it was initially released. However, only a couple of weeks later, Google released "Imagen" which eclipsed DALL E's results.
More recently, another algorithm called "Stable Diffusion" has been released, which is on-par with DALL E and Imagen, but is completely open-source. Stable Diffusion is available on NightCafe along with other (more primitive) generators, all of which will create images based on your prompts and are worth playing around with.
Why is it Called DALL-E?
The name DALL-E is a combination of WALL-E (the Pixar movie) and Salvador Dalí the artist.
DALL·E 2 Content Policy
Use of the DALL-E 2 algorithm is subject to OpenAI's Content Policy. Unlike our other algorithms, creations must be G-rated, and you are not allowed to create images of celebrities or public figures.
Please read the full OpenAI Content Policy (it's really not very long).