With all the talk about AI replacing writers, here's an app that could work to bring out our inner artists. Have not tried it myself, but love the idea of using words to instruct the app on what to draw. I'm thinking logos and graphics, or anyone who wants to create unique cover art for self-publishing. (Apologies to professional artists!)
Copied from an article in the December 2022 edition of Swissquote, a business magazine. (https://en.swissquote.com/sites/default/files/2022-12/sq6_en_epaper.pdf)
BECOME A PROFESSIONAL ARTIST WITH AI
BY GÉRARD DUCLOS
With all the progress made in the world of AI, now anyone can instantly become an accomplished artist. We tested this fascinating technology with the MidJourney program.
In January 2021, at the height of the pandemic, OpenAI, an AI research lab that brings together the best of US tech companies, announced the launch of DALL-E, a program that creates images from text descriptions.
The concept involves users typing words or phrases and the program creating images based on those words. The more precise the words, the better the images turn out.
This was quickly followed by DALL-E 2, an improved version, as well as competitor programs. This breakthrough marks the start of mass-market availability of artificial intelligence that has been developed for years in Silicon Valley labs and trained for more than a decade by billions of CAPTCHAs completed for free by billions of users (AI thanks you).
A competitor of DALL-E, MidJourney stands out for its user-friendly approach and users can generate up to 25 free images before switching to the subscription option. To start the process, we connected to the Discord server to supply the program with our text.
The first surprise is that there is no predefined list of keywords or commands. You can use whatever combination of words you like, even the most absurd or far-fetched phrases, such as “a trader in a red suit sitting in a chair reads a financial magazine while a screen behind him shows the market crashing to 0”. Hmm, not a bad idea, let’s try that sentence! I submit the request and the program generates four iterations of the image in just a few seconds. The result is incredible. AI actually drew four traders in red suits reading a financial magazine while a screen behind them shows the market crashing, and it is actually rather pleasing to look at, even if there are a few imperfections.
Even though there isn’t a list of keywords per se, we can give the program all sorts of details about the style and quality that we are expecting from the generated image: super realistic, cyberpunk, in the style of Rembrandt, 8k, incredibly detailed, etc. The possibilities are endless we soon realise, as we admire drawings done by much more advanced users who are using the program at the same
Some of the drawings are so impressive that it is virtually impossible to believe that the final image was drawn entirely by artificial intelligence. The Discord server is full of fascinating examples of high-quality images, such as entire comic books drawn by AI in a coherent and very aesthetically pleasing manner.
For more abstract but just as surprising results, you can also try short combinations of words or abstract concepts and let yourself be amazed by the program. For example, we tried “sunset on medieval Switzerland”. Once again, the result is incredibly beautiful and detailed.
At the end of our test, we were amazed and terrified in equal measure. One thing is certain: AI is on track to replace humans in applications which we thought it could never do, such as art, design and aestheticism... And it will only get better in the future.
Copied from an article in the December 2022 edition of Swissquote, a business magazine. (https://en.swissquote.com/sites/default/files/2022-12/sq6_en_epaper.pdf)
BECOME A PROFESSIONAL ARTIST WITH AI
BY GÉRARD DUCLOS
With all the progress made in the world of AI, now anyone can instantly become an accomplished artist. We tested this fascinating technology with the MidJourney program.
In January 2021, at the height of the pandemic, OpenAI, an AI research lab that brings together the best of US tech companies, announced the launch of DALL-E, a program that creates images from text descriptions.
The concept involves users typing words or phrases and the program creating images based on those words. The more precise the words, the better the images turn out.
This was quickly followed by DALL-E 2, an improved version, as well as competitor programs. This breakthrough marks the start of mass-market availability of artificial intelligence that has been developed for years in Silicon Valley labs and trained for more than a decade by billions of CAPTCHAs completed for free by billions of users (AI thanks you).
A competitor of DALL-E, MidJourney stands out for its user-friendly approach and users can generate up to 25 free images before switching to the subscription option. To start the process, we connected to the Discord server to supply the program with our text.
The first surprise is that there is no predefined list of keywords or commands. You can use whatever combination of words you like, even the most absurd or far-fetched phrases, such as “a trader in a red suit sitting in a chair reads a financial magazine while a screen behind him shows the market crashing to 0”. Hmm, not a bad idea, let’s try that sentence! I submit the request and the program generates four iterations of the image in just a few seconds. The result is incredible. AI actually drew four traders in red suits reading a financial magazine while a screen behind them shows the market crashing, and it is actually rather pleasing to look at, even if there are a few imperfections.
Even though there isn’t a list of keywords per se, we can give the program all sorts of details about the style and quality that we are expecting from the generated image: super realistic, cyberpunk, in the style of Rembrandt, 8k, incredibly detailed, etc. The possibilities are endless we soon realise, as we admire drawings done by much more advanced users who are using the program at the same
Some of the drawings are so impressive that it is virtually impossible to believe that the final image was drawn entirely by artificial intelligence. The Discord server is full of fascinating examples of high-quality images, such as entire comic books drawn by AI in a coherent and very aesthetically pleasing manner.
For more abstract but just as surprising results, you can also try short combinations of words or abstract concepts and let yourself be amazed by the program. For example, we tried “sunset on medieval Switzerland”. Once again, the result is incredibly beautiful and detailed.
At the end of our test, we were amazed and terrified in equal measure. One thing is certain: AI is on track to replace humans in applications which we thought it could never do, such as art, design and aestheticism... And it will only get better in the future.