See more photos…
Butterflies on Wildflowers Art by Famous Artists Via Dall-E 2 by Ronald Horii
While playing with OpenAI's ChatGPT, I noticed they had another program called DALL-E 2: https://openai.com/dall-e-2/. While ChatGPT is a smart text generator, DALL-E 2 is an AI application program that can generate images based on a description that you enter, up to 400 characters. I decided to try it out by giving it one subject: "Butterflies on wildflowers," and asked it to generate images in the styles of different artists. Here's what it came up with. It generated 4 images for each artist. I'm not an art critic, so I can't comment on the artistic merits of the results, but they do show elements of the artists' style, some better than others. The pictures at the end are not according to the style of any particular artist.
DALL-E 2 is free to use for a limited number of images. (You get 50 free credits for the first month, then 15 per month after that. It takes 1 credit per image request, which gives you 4 images.) You can buy credits to get more images. ($15 for 115 credits.) Here's a tutorial on how to use it: https://www.youtube.com/watch?v=bEq7NfNH58Q. This addresses the ethical questions about using DALL-E 2 to generate art: https://www.youtube.com/watch?v=-W3LQhM5tMk. Here are reviews of DALL-E 2: https://www.dpreview.com/news/7603049074/incredible-ai-system-dall-e-2-enters-public-beta, https://beincrypto.com/learn/dall-e-2-review/, https://www.zdnet.com/article/how-to-use-dalloe-2-to-turn-your-wildest-imaginations-into-ai-generated-art/ (also reviews other AI art generators). Here are reviews of other AI art generators, along with DALL-E 2: https://www.unite.ai/10-best-ai-art-generators/.
How does DALL-E 2 work? I asked ChatGPT for an explanation. Here's what it said: "DALL-E 2 is a generative model developed by OpenAI that creates images from textual descriptions. The model is trained on a massive dataset of text and image pairs, which allows it to learn the relationship between textual descriptions and corresponding visual concepts. During inference, given a textual description, the model generates an image that corresponds to the description by using a decoder to transform the textual input into a high-dimensional representation, which is then passed through a series of layers to generate the final image." Where does DALL-E 2 get its source images?
"DALL-E 2, the artificial intelligence model developed by OpenAI, generates images from textual descriptions by using a large-scale generative model trained on diverse Internet images. These images come from a variety of sources, including search engines, social media, and online archives, among others. The model has been trained on a diverse and extensive dataset to ensure that it can generate a wide range of images from a diverse set of textual descriptions." Does DALL-E 2 modify images after it find them?
"DALL-E 2 does not simply modify existing images; it generates new images from scratch based on the textual description provided as input. The model has been trained on a massive dataset of diverse images and uses that training to generate novel images that match the description given. The output is a completely new and unique image that has never existed before." This is a technical paper on how it works: https://arxiv.org/abs/2204.06125.
While playing with OpenAI's ChatGPT, I noticed they had another program called DALL-E 2: https://openai.com/dall-e-2/. While ChatGPT is a smart text generator, DALL-E 2 is an AI application program that can generate images based on a description that you enter, up to 400 characters. I decided to try it out by giving it one subject: "Butterflies on wildflowers," and asked it to generate images in the styles of different artists. Here's what it came up with. It generated 4 images for each artist. I'm not an art critic, so I can't comment on the artistic merits of the results, but they do show elements of the artists' style, some better than others. The pictures at the end are not according to the style of any particular artist.
DALL-E 2 is free to use for a limited number of images. (You get 50 free credits for the first month, then 15 per month after that. It takes 1 credit per image request, which gives you 4 images.) You can buy credits to get more images. ($15 for 115 credits.) Here's a tutorial on how to use it: https://www.youtube.com/watch?v=bEq7NfNH58Q. This addresses the ethical questions about using DALL-E 2 to generate art: https://www.youtube.com/watch?v=-W3LQhM5tMk. Here are reviews of DALL-E 2: https://www.dpreview.com/news/7603049074/incredible-ai-system-dall-e-2-enters-public-beta, https://beincrypto.com/learn/dall-e-2-review/, https://www.zdnet.com/article/how-to-use-dalloe-2-to-turn-your-wildest-imaginations-into-ai-generated-art/ (also reviews other AI art generators). Here are reviews of other AI art generators, along with DALL-E 2: https://www.unite.ai/10-best-ai-art-generators/.
How does DALL-E 2 work? I asked ChatGPT for an explanation. Here's what it said: "DALL-E 2 is a generative model developed by OpenAI that creates images from textual descriptions. The model is trained on a massive dataset of text and image pairs, which allows it to learn the relationship between textual descriptions and corresponding visual concepts. During inference, given a textual description, the model generates an image that corresponds to the description by using a decoder to transform the textual input into a high-dimensional representation, which is then passed through a series of layers to generate the final image." Where does DALL-E 2 get its source images?
"DALL-E 2, the artificial intelligence model developed by OpenAI, generates images from textual descriptions by using a large-scale generative model trained on diverse Internet images. These images come from a variety of sources, including search engines, social media, and online archives, among others. The model has been trained on a diverse and extensive dataset to ensure that it can generate a wide range of images from a diverse set of textual descriptions." Does DALL-E 2 modify images after it find them?
"DALL-E 2 does not simply modify existing images; it generates new images from scratch based on the textual description provided as input. The model has been trained on a massive dataset of diverse images and uses that training to generate novel images that match the description given. The output is a completely new and unique image that has never existed before." This is a technical paper on how it works: https://arxiv.org/abs/2204.06125.
1 Like
Judy Gallagher likes this.
