<?xml version="1.0" encoding="UTF-8"?><rss version="2.0"
	xmlns:content="http://purl.org/rss/1.0/modules/content/"
	xmlns:dc="http://purl.org/dc/elements/1.1/"
	xmlns:atom="http://www.w3.org/2005/Atom"
	xmlns:sy="http://purl.org/rss/1.0/modules/syndication/"
	
	>
<channel>
	<title>
	Comments on: Deconstructing AI Art masterpieces: Create Art like Refik Anadol	</title>
	<atom:link href="https://manfred-feiger.com/create-ai-art-like-refik-anadol/feed/" rel="self" type="application/rss+xml" />
	<link>https://manfred-feiger.com/create-ai-art-like-refik-anadol/</link>
	<description>Your Swiss-army knife for your next digital project. Let&#039;s meet and write our story together.</description>
	<lastBuildDate>Wed, 08 Nov 2023 06:54:45 +0000</lastBuildDate>
	<sy:updatePeriod>
	hourly	</sy:updatePeriod>
	<sy:updateFrequency>
	1	</sy:updateFrequency>
	
<atom:link rel="hub" href="https://pubsubhubbub.appspot.com"/>
<atom:link rel="hub" href="https://pubsubhubbub.superfeedr.com"/>
<atom:link rel="hub" href="https://websubhub.com/hub"/>
<atom:link rel="self" href="https://manfred-feiger.com/create-ai-art-like-refik-anadol/feed/"/>
	<item>
		<title>
		By: Craig		</title>
		<link>https://manfred-feiger.com/create-ai-art-like-refik-anadol/#comment-1194</link>

		<dc:creator><![CDATA[Craig]]></dc:creator>
		<pubDate>Mon, 13 Mar 2023 17:35:45 +0000</pubDate>
		<guid isPermaLink="false">https://manfred-feiger.com/?p=940#comment-1194</guid>

					<description><![CDATA[Thanks so much for this collection of links. Very helpful.]]></description>
			<content:encoded><![CDATA[<p>Thanks so much for this collection of links. Very helpful.</p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		By: Generative Art – how to get started with AI Art - Manfred Feiger		</title>
		<link>https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/</link>

		<dc:creator><![CDATA[Generative Art – how to get started with AI Art - Manfred Feiger]]></dc:creator>
		<pubDate>Tue, 27 Dec 2022 15:31:35 +0000</pubDate>
		<guid isPermaLink="false">https://manfred-feiger.com/?p=940#comment-32</guid>

					<description><![CDATA[This Article was mentioned on &lt;a href=&quot;https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/&quot;&gt;manfred-feiger.com&lt;/a&gt;]]></description>
			<content:encoded><![CDATA[<p>This Article was mentioned on <a href="https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/">manfred-feiger.com</a></p>
]]></content:encoded>
		
			</item>
		<item>
		<title>
		By: manfred feiger		</title>
		<link>https://manfred-feiger.com/speaking-computer-language-welcome-ai-art/Speaking%20computer%20language%20–%20welcome%20AI%20Art</link>

		<dc:creator><![CDATA[manfred feiger]]></dc:creator>
		<pubDate>Tue, 17 May 2022 13:18:29 +0000</pubDate>
		<guid isPermaLink="false">https://manfred-feiger.com/?p=940#comment-768</guid>

					<description><![CDATA[&lt;a href=&quot;https://manfred-feiger.com/speaking-computer-language-welcome-ai-art/&quot; rel=&quot;ugc&quot;&gt;Speaking computer language – welcome AI Art&lt;/a&gt;&lt;b&gt;Published:&lt;/b&gt; May 17, 2022&lt;strong&gt;Reading time&lt;/strong&gt; &#060; 12 minutes&lt;b&gt;Categories:&lt;/b&gt; &lt;a href=&quot;https://manfred-feiger.com/category/art/&quot; rel=&quot;ugc&quot;&gt;art&lt;/a&gt; &#124; &lt;a href=&quot;https://manfred-feiger.com/category/design/&quot; rel=&quot;ugc&quot;&gt;design&lt;/a&gt; &#124; &lt;a href=&quot;https://manfred-feiger.com/category/future/&quot; rel=&quot;ugc&quot;&gt;future&lt;/a&gt; &#124; &lt;a href=&quot;https://manfred-feiger.com/category/ml/&quot; rel=&quot;ugc&quot;&gt;ML&lt;/a&gt;&lt;b&gt;by:&lt;/b&gt; &lt;a href=&quot;https://manfred-feiger.com&quot; rel=&quot;ugc&quot;&gt;manfred feiger&lt;/a&gt;2022-05-17T13:18:29+00:00Natural language as an input for creating AI art changes the way we sketch ideas. DALL·E 2 or Imagen give us a glimpse into AI graphics.

The command prompt: A black, empty screen and a blinking line, waiting for my input. Could you imagine that this prompt is the access to art or AI art? The command line shows a visual result of your terms, not ASCII art or related items. Nowadays, those prompts have different shapes, such as a form field or the use of voice interfaces. 
Your imagination of thought, translated to words in your natural language (mostly in English), gives you access to new creativity. In the words of advertising, you need to express a good core idea, you want to get visualized. Similar to creatives being used to Image search engines, the need to find the correct words gets more and more important.
To explore my thoughts I look at the evolution of the way we tell machines, what to do and how CLIP and other neural models transform the way we could create what we are looking for. Whether it be AI art or maybe any picture we want to see, in general or what I would call AI graphics.
Computer language
A computer language is a language we use to talk to machines. It has been in use for about 70 years.In terms of human history, a very short time.
The first, more popular programming languages found their roots between 1960 and 1970. The first “high-level programming language” was &lt;a href=&quot;https://en.wikipedia.org/wiki/Plankalk%C3%BCl&quot; rel=&quot;nofollow ugc&quot;&gt;Plankalkül&lt;/a&gt; from around 1945. Even earlier, the roots of programming go back to Ada Lovelace, yes, a woman, in the year 1883.
Nowadays, we even have low code or no-code ways to speak with machines. Mostly those use visual environments to drag together code components or nodes and are called visual programming languages.
So, in the past, we needed to write code in a language, following grammar. Now, we could also drag together many nodes and hope for an effect, while playing with parameters. At some point of customization those visual environments also allow to add code manually. But it&#039;s not essentially necessary to write the code yourself anymore.
DALL·E and other neural networks allow access to Artificial intelligence as a source of processing your thoughts via a text prompt. If you want, you could say that you tell the machine, what it should program with your natural language. Your chosen words define the quality of the output you are aiming for.
Natural language and natural interfaces
The natural language is the way we as humans speak with each other. To some extent even our voice-interfaces around us are now able to communicate with us in a way that feels almost natural. In this case our command prompt mostly starts with “Alexa”, “Siri” or “Hey Google” to initialize the dialogue. Natural language processing takes the task of converting speech into words in the background to take the right commands.
In 2013, there was a trend to use natural language forms in interfaces to make form usage feel more natural. Forms were integrated into sentences rather than presented as a separate option. Stats are more usable in natural language, but confusing numbers make it difficult to understand the impact of improvements.
The tendency to make interfaces more natural or more human, so we could use our natural senses to interact, is present. But the evolution of the interfaces also depends on technological developments.
&lt;a href=&quot;https://vimeo.com/420773167?embedded=true&#038;source=vimeo_logo&#038;owner=53957807&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;/a&gt;Venous materials from MIT lab – Watch example on Vimeo
There are many interesting developments, especially if you investigate the MIT Media lab and the tangible media group. They strive for seamlessly coupling the worlds of bits and atoms by giving dynamic physical form to digital information and computation. &lt;a href=&quot;https://www.media.mit.edu/projects/prototyping-interactive-fluidic-mechanisms/overview/&quot; rel=&quot;nofollow ugc&quot;&gt;Research on venous materials and human-flora interaction provides insights into our future.&lt;/a&gt;
One of the easiest ways to interact now is our language, expressed by voice interfaces. When using Neuralink or similar chips inside our bodies, it&#039;s likely that we&#039;ll need to use interfaces with a well-structured language initially.
If you are looking for more scientific terms on how the language is being used for generating visual results, look for &lt;a href=&quot;https://www.google.com/search?q=text-to-image+synthesis&#038;rlz=1C5CHFA_enDE985DE985&#038;oq=text-to-image+synthesis&#038;aqs=chrome..69i57j0i19j0i19i22i30l5j69i61.1431j0j1&#038;sourceid=chrome&#038;ie=UTF-8&quot; rel=&quot;nofollow ugc&quot;&gt;text-to-image synthesis&lt;/a&gt;.
GANs produce visual effects resembling the WebGL example on alteredqualia.com that utilizes DeepWarp, a method commonly employed for gaze movement.

CLIP, GANs or AI driven Art and approaches of deep learning.
Natural language processing is also an important tool when it comes to playing with so called CLIPs. CLIP is “Contrast Language-Image Pre-Training”, a neural network trained on a variety of text-image pairs. CLIP got introduced around January 2021.
Before getting into CLIP, a short excursion on learning. Deep learning is part of machine learning and helps machines to learn. The machine is supposed to increase its own skills without the help of humans. On contrast to machine learning, where humans interfere in the learning process, in deep learning humans take care of providing information for learning. To do so, we could also use our input prompt. The process of analysis is up to the machine itself.
An example for machine learning would be connecting words and images by labelling the images with the appropriate word. A typical combination for this cause would be &lt;a href=&quot;https://www.image-net.org/&quot; rel=&quot;nofollow ugc&quot;&gt;ImageNet&lt;/a&gt; and &lt;a href=&quot;https://wordnet.princeton.edu/&quot; rel=&quot;nofollow ugc&quot;&gt;WordNet&lt;/a&gt;.
CLIP is a neural network model that is already trained on 400,000,000 (image, text) pairs. So, this means that this is the data that is used in training the CLIP Model. You can input an image to the model and get the most probable caption or summary of it. Alternatively, you can use natural language as an input. The true magic comes when you see examples. &lt;a href=&quot;https://twitter.com/metasemantic&quot; rel=&quot;nofollow ugc&quot;&gt;Travis &lt;/a&gt;&lt;a href=&quot;https://twitter.com/metasemantic&quot; rel=&quot;nofollow ugc&quot;&gt;Hoppe&lt;/a&gt; shares some interesting applications in his Twitter account.
To get a full “beginners” understanding of the CLIP model, I recommend this article: “&lt;a href=&quot;https://www.kdnuggets.com/2021/03/beginners-guide-clip-model.html&quot; rel=&quot;nofollow ugc&quot;&gt;A beginner’s guide to the CLIP Model&lt;/a&gt;”. You will find many interesting articles on &lt;a href=&quot;http://kdnuggets.com&quot; rel=&quot;nofollow ugc&quot;&gt;kdnuggets.com&lt;/a&gt; around artificial intelligence, computer vision, data science, machine learning, natural language processing or programming (Python).
To try CLIP yourself, you could use &lt;a href=&quot;https://compvis.github.io/taming-transformers/&quot; rel=&quot;nofollow ugc&quot;&gt;VQGAN&lt;/a&gt; developed by the University of Heidelberg. It’s a GAN (a generative adversarial network), those are special types of neural networks consisting of “Generator” and “discriminator” networks. There is a GAN that generates images. The generator is like an artist and the discriminator is like a critic during training. When the generator can produce an image that the discriminator can&#039;t distinguish from a human-made one, we can use it for creating art.
There are huge libraries of available GANs for very different purposes. Some of the lists I could recommend:
&lt;a href=&quot;https://github.com/hindupuravinash/the-gan-zoo&quot; rel=&quot;nofollow ugc&quot;&gt;The GAN Zoo&lt;/a&gt;
&lt;a href=&quot;https://github.com/nashory/gans-awesome-applications&quot; rel=&quot;nofollow ugc&quot;&gt;GANS awesome applications&lt;/a&gt;
To play with the technology to learn, use &lt;a href=&quot;https://www.artbreeder.com/&quot; rel=&quot;nofollow ugc&quot;&gt;artbreeder&lt;/a&gt; and generate new images; Artbreeder is also based on GAN based image generation.

&lt;a href=&quot;https://www.artbreeder.com/&quot; rel=&quot;nofollow ugc&quot;&gt;Artbreeder&lt;/a&gt;, Image generation based on GAN
Exploring Applications of CLIP
How to get started? I would always recommend playing with the technology. Most technology is too complex for an early deep dive, so start by playing. If you found your motivation to dig deeper, do it with all your might as you need a lot of patience to get a good understanding.
So, let’s look at &lt;a href=&quot;https://colab.research.google.com/github/justinjohn0306/VQGAN-CLIP/blob/main/VQGAN%2BCLIP(Updated).ipynb#scrollTo=JqFEaEChR20v&quot; rel=&quot;nofollow ugc&quot;&gt;VQGAN + CLIP&lt;/a&gt; you could use directly yourself online. This is a collab notebook, a small development environment running Python programming language. Your task is to input some values... To get started, I would recommend to only change the “natural language prompt” of what you want to see. In my case I tested some results of different apps with the sentence “Children playing in the forest” (because my kids are attending a kindergarten in the wood while writing these lines).
These are my first results using the CLIP linked above (to try yourself, open the &lt;a href=&quot;https://colab.research.google.com/github/justinjohn0306/VQGAN-CLIP/blob/main/VQGAN%2BCLIP(Updated).ipynb&quot; rel=&quot;nofollow ugc&quot;&gt;link&lt;/a&gt;, basically go through the page and always execute the play button to the left and wait till there’s a check icon next to it, to signal it’s loaded), only for chapter “Parameters” enter a text. In the end you could also create a video. The results here are early iterations. The program would continue till the “free” volume is used up.

First iterations with VQGAN + CLIP
Number 100
Number 150


from &lt;a href=&quot;http://wombo.art&quot; rel=&quot;nofollow ugc&quot;&gt;wombo.art&lt;/a&gt;
from &lt;a href=&quot;https://creator.nightcafe.studio/&quot; rel=&quot;nofollow ugc&quot;&gt;nightcafe&lt;/a&gt;

To compare my result, I checked for other applications out there and their result. So there’s for example an application that is also available as an app; In &lt;a href=&quot;http://wombo.art&quot; rel=&quot;nofollow ugc&quot;&gt;wombo.art&lt;/a&gt; you could test directly. The result from wombo art had similar patterns as my test with YQGAN and looked like the picture on the at the top left.
Then I used another popular web-service to compare: &lt;a href=&quot;https://creator.nightcafe.studio/&quot; rel=&quot;nofollow ugc&quot;&gt;Nightcafe&lt;/a&gt; you can see on the right.
You see, when using a webservice, you end up getting one result. Using a &lt;a href=&quot;https://colab.research.google.com/&quot; rel=&quot;nofollow ugc&quot;&gt;google colab&lt;/a&gt;, you get many iterations and it&#039;s not as fast as the webservice or app. (Google Colab or Colaboratory is a way to execute Pyhton code in the browser).
There are also other CLIPS on google colab you could use. The following picture is the (an early iterative) result of the combination &lt;a href=&quot;https://colab.research.google.com/drive/1L8oL-vLJXVcRzCFbPwOoMkPKJ8-aYdPN#scrollTo=g7EDme5RYCrt&quot; rel=&quot;nofollow ugc&quot;&gt;VQGAN+CLIP (z+quantize method)&lt;/a&gt;:

You could continue playing with many more version out there. Here you will find an overview of newer applications: &lt;a href=&quot;https://softologyblog.wordpress.com/&quot; rel=&quot;nofollow ugc&quot;&gt;https://softologyblog.wordpress.com/&lt;/a&gt;
You notice that the basic style follows one pattern and it’s obvious nightcafe also uses some way of VQGAN combination. To create magic, or what I would call magic based on a text prompt, there’s still some way to go.
&lt;strong&gt;The magic I aim for: I tell the machine what I want to see, and they give me iterations of the thing I type in in. It mustn&#039;t be art or AI art&lt;/strong&gt;.
To create this magic, we look at DALL·E. &lt;a href=&quot;https://openai.com/&quot; rel=&quot;nofollow ugc&quot;&gt;OpenAI&lt;/a&gt; released DALL·E 2 in April 2022. They show the real magic of the future. By the way, I wouldn&#039;t call the results art. Mostly we generate graphics.
DALL·E 2 screenshot, see more on their &lt;a href=&quot;https://openai.com/dall-e-2/&quot; rel=&quot;nofollow ugc&quot;&gt;website&lt;/a&gt;
A short introduction to DALL·E, DALL·E 2 and DALL·E mini
&lt;em&gt;“DALL·E 2 is an artificial intelligence model that takes a text prompt and/or existing image as an input and generates a new image as an output. DALL·E 2 was developed by researchers at OpenAI to understand the capabilities and broader implications of multimodal generative models. To help us and others better understand how image generation models can be used and misused.”&lt;/em&gt; (From their own description)
To make it shot, look at &lt;a href=&quot;https://openai.com/dall-e-2/&quot; rel=&quot;nofollow ugc&quot;&gt;their website&lt;/a&gt; and get a feeling of the beauty that could be generated with DALL·E 2 or watch the &lt;a href=&quot;https://www.youtube.com/watch?v=qTgPSKKjfVg&#038;t=93s&quot; rel=&quot;nofollow ugc&quot;&gt;short video&lt;/a&gt; on youtube:
&lt;a href=&quot;https://www.youtube.com/watch?v=qTgPSKKjfVg&#038;t=93s&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;/a&gt;
Unfortunately, DALL·E 2 is not easily available to test yet. But you could sign up on their waiting list.
According to OpenAI DALL·E 2 isn&#039;t trained on faces, to avoid the risk of deep fakes. The danger is already there. Finally it the time to doubt any images being shown to us!
On twitter try a search for &lt;a href=&quot;https://twitter.com/hashtag/Dalle?src=hashtag_click&quot; rel=&quot;nofollow ugc&quot;&gt;#dalle&lt;/a&gt; or &lt;a href=&quot;https://twitter.com/hashtag/dalle2?src=hashtag_click&quot; rel=&quot;nofollow ugc&quot;&gt;#&lt;/a&gt;&lt;a href=&quot;https://twitter.com/hashtag/dalle2?src=hashtag_click&quot; rel=&quot;nofollow ugc&quot;&gt;dalle2&lt;/a&gt; and you find many inspiring pictures. Going through the magical results, it feels as if image searches could get a relic of the past soon or will only be needed for special causes. Image stock market could be AI generated in my point of view. Here are some randomly selected shots from some twitter members (all linked):
&lt;a href=&quot;https://twitter.com/hardmaru/status/1524303013766373377&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;/a&gt;
&lt;a href=&quot;https://twitter.com/djbaskin_images/status/1523816614713470976&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;/a&gt;
&lt;a href=&quot;https://twitter.com/hardmaru/status/1523664409838510082?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1523664409838510082%7Ctwgr%5E%7Ctwcon%5Es1_&#038;ref_url=https%3A%2F%2Fpublish.twitter.com%2F%3Fquery%3Dhttps3A2F2Ftwitter.com2Fhardmaru2Fstatus2F1523664409838510082widget%3DTweet&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;/a&gt;
Yes, it’s a pity, we all can’t directly play with this fascinating technology!
Luckily I found &lt;a href=&quot;https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;strong&gt;DALL·E mini&lt;/strong&gt;&lt;/a&gt;&lt;strong&gt;.&lt;/strong&gt; And that&#039;s your friend to play! By the way, I also created my post image with Dall-e mini. I asked for &quot;AI Art blue and yellow&quot;. One of the results could be found on the overview site of my &lt;a href=&quot;https://manfred-feiger.com/writing/&quot; rel=&quot;nofollow ugc&quot;&gt;writings&lt;/a&gt;.
The description of &lt;a href=&quot;https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;strong&gt;DALL·E mini&lt;/strong&gt;&lt;/a&gt;: &lt;em&gt;“OpenAI had the first impressive model for generating images through DALL·E. DALL·E mini is an attempt at reproducing those results with an open-source model.”&lt;/em&gt;
They also have DALL·E mega, a large version of the mini, that is being trained. To see the fascinating world of training, look into the &lt;a href=&quot;https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-Mega-Training--VmlldzoxODMxMDI2&quot; rel=&quot;nofollow ugc&quot;&gt;training space&lt;/a&gt;.
And finally, here’s your playground to play with the DALL-E mini: &lt;a href=&quot;https://huggingface.co/spaces/dalle-mini/dalle-mini&quot; rel=&quot;nofollow ugc&quot;&gt;https://huggingface.co/spaces/dalle-mini/dalle-mini&lt;/a&gt;
That’s my result on &quot;playing children in the forest&quot;; quite an improvement in a very short time. If you play more with &lt;a href=&quot;https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy&quot; rel=&quot;nofollow ugc&quot;&gt;&lt;strong&gt;DALL·E mini&lt;/strong&gt;&lt;/a&gt; you will notice, that faces still are really not trained much. But I already love it!

To feel the new power of words and to discover some magic along using words, go out and play for yourself!
By the way, there&#039;s another tool on the web, giving you similar results. It&#039;s &lt;a href=&quot;https://neuralblender.com/&quot; rel=&quot;nofollow ugc&quot;&gt;neuralblender&lt;/a&gt;. Here&#039;s my kids results from them:

If you want to get inspired by other AI Artists, look at the work of &lt;a href=&quot;https://manfred-feiger.com/create-ai-art-like-refik-anadol/&quot; rel=&quot;nofollow ugc&quot;&gt;Refik Anadol&lt;/a&gt;, his process is the application of a lot of machine learning to a real world of art. As I said, in my point of view, the term of AI Art is mostly wrong, looking at the output of DALL·E or other CLIP applications, as the outputs are not expressive. In my point of view, we speak about AI graphics.
Maybe you could guess, what my prompt in this case was (created with DALL·E Mini):


&lt;strong&gt;Have fun playing!&lt;/strong&gt;
And guess what... My challenge is to build websites via a text prompt. Currently I am looking for available sources to have a good segmentation (Structural attributes, Style attributes). Let&#039;s see when I will share my results or findings.

Some more tipps on playing with &lt;a href=&quot;https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy&quot; rel=&quot;nofollow ugc&quot;&gt;DALL·E mini&lt;/a&gt;... use objects. To avoid deep fakes, faces won&#039;t work. Play around with the text prompt. You could try with the attribute, in the style of... or try the &lt;a href=&quot;https://docs.google.com/spreadsheets/d/1y7nAbmR4FREi6npB1u-Bo3GFdwdOPYJc617rBOxIRHY/edit#gid=0&quot; rel=&quot;nofollow ugc&quot;&gt;DrawBench&lt;/a&gt; prompts being used for Google Imagen first.
Dalle Mini: a rocket in the style of Matsuo Basho

Google Imagen
While publishing this article on Mai 24th, Google came up with a competitor to OpenAI DALL·E 2. The market is on fire! And Google-&lt;a href=&quot;https://imagen.research.google/&quot; rel=&quot;nofollow ugc&quot;&gt;Imagen&lt;/a&gt; impresses the same way, DALL·E 2 did. 
Be excited to check out the first tweets on the topic &lt;a href=&quot;https://twitter.com/hashtag/imagen?src=hashtag_click&quot; rel=&quot;nofollow ugc&quot;&gt;#imagen&lt;/a&gt;.
And we will all be happy to play with the first, real demos!
Screenshot from &lt;a href=&quot;https://imagen.research.google/&quot; rel=&quot;nofollow ugc&quot;&gt;Imagen&lt;/a&gt; website

Recommended further readings

&lt;a href=&quot;https://mirror.xyz/herndondryhurst.eth&quot; rel=&quot;nofollow ugc&quot;&gt;Infinite Images and the latent camera&lt;/a&gt;
&lt;a href=&quot;https://arxiv.org/abs/2204.06125&quot; rel=&quot;nofollow ugc&quot;&gt;Hierarchical Text-Conditional Image Generation with CLIP Latents&lt;/a&gt;
&lt;a href=&quot;https://openai.com/dall-e-2/&quot; rel=&quot;nofollow ugc&quot;&gt;DALL·E 2&lt;/a&gt;
&lt;a href=&quot;https://imagen.research.google/&quot; rel=&quot;nofollow ugc&quot;&gt;Imagen by Google&lt;/a&gt;
&lt;a href=&quot;https://nvlabs.github.io/stylegan3/&quot; rel=&quot;nofollow ugc&quot;&gt;StyleGan 3&lt;/a&gt;
Explore my newer articles around the topics:&lt;a href=&quot;https://manfred-feiger.com/how-ai-assistants-help-you-with-text-and-art/&quot; rel=&quot;ugc&quot;&gt;How AI Assistants Can Help You with Text and Art&lt;/a&gt;&lt;a href=&quot;https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/&quot; rel=&quot;ugc&quot;&gt;Generative Art – how to get started with AI Art&lt;/a&gt;&lt;a href=&quot;https://manfred-feiger.com/midjourney-and-dall-e-2-first-hands-on-impressions/&quot; rel=&quot;ugc&quot;&gt;Midjourney and DALL-E 2 – first hands on impressions&lt;/a&gt;&lt;a href=&quot;https://manfred-feiger.com/how-stable-diffusion-celebrates-new-forms-of-creativity/&quot; rel=&quot;ugc&quot;&gt;Stable Diffusion celebrates new forms of creativity&lt;/a&gt;


FAQs
What are AI GraphicsVisual images or designs on some surface, such as a wall, canvas, screen, paper, or stone, to inform, illustrate, or entertain being generated with the help of Artificial Intelligence.
&lt;a href=&quot;https://manfred-feiger.com/create-ai-art-like-refik-anadol/&quot; rel=&quot;ugc&quot;&gt;Deconstructing AI Art masterpieces: Create Art like Refik Anadol&lt;/a&gt;What If you were asked to create art like Refik Anadol? Or to produce some &quot;AI art&quot;?
In case you
&lt;a href=&quot;https://manfred-feiger.com/midjourney-and-dall-e-2-first-hands-on-impressions/&quot; rel=&quot;ugc&quot;&gt;Midjourney and DALL-E 2 – first hands on impressions&lt;/a&gt;Maybe you remember my article speaking computer language – welcome AI art, speaking about the new role of t
&lt;a href=&quot;https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/&quot; rel=&quot;ugc&quot;&gt;Generative Art – how to get started with AI Art&lt;/a&gt;The world of art is a diverse and beautiful place. The more we explore new forms of art, the more we are
&lt;a href=&quot;https://manfred-feiger.com/how-ai-assistants-help-you-with-text-and-art/&quot; rel=&quot;ugc&quot;&gt;How AI Assistants Can Help You with Text and Art&lt;/a&gt;Have you ever wished you had a personal assistant to help you with your writing or with your artworks?
AI]]></description>
			<content:encoded><![CDATA[<p><a href="https://manfred-feiger.com/speaking-computer-language-welcome-ai-art/" rel="ugc">Speaking computer language – welcome AI Art</a><b>Published:</b> May 17, 2022<strong>Reading time</strong> &lt; 12 minutes<b>Categories:</b> <a href="https://manfred-feiger.com/category/art/" rel="ugc">art</a> | <a href="https://manfred-feiger.com/category/design/" rel="ugc">design</a> | <a href="https://manfred-feiger.com/category/future/" rel="ugc">future</a> | <a href="https://manfred-feiger.com/category/ml/" rel="ugc">ML</a><b>by:</b> <a href="https://manfred-feiger.com" rel="ugc">manfred feiger</a>2022-05-17T13:18:29+00:00Natural language as an input for creating AI art changes the way we sketch ideas. DALL·E 2 or Imagen give us a glimpse into AI graphics.</p>
<p>The command prompt: A black, empty screen and a blinking line, waiting for my input. Could you imagine that this prompt is the access to art or AI art? The command line shows a visual result of your terms, not ASCII art or related items. Nowadays, those prompts have different shapes, such as a form field or the use of voice interfaces.<br />
Your imagination of thought, translated to words in your natural language (mostly in English), gives you access to new creativity. In the words of advertising, you need to express a good core idea, you want to get visualized. Similar to creatives being used to Image search engines, the need to find the correct words gets more and more important.<br />
To explore my thoughts I look at the evolution of the way we tell machines, what to do and how CLIP and other neural models transform the way we could create what we are looking for. Whether it be AI art or maybe any picture we want to see, in general or what I would call AI graphics.<br />
Computer language<br />
A computer language is a language we use to talk to machines. It has been in use for about 70 years.In terms of human history, a very short time.<br />
The first, more popular programming languages found their roots between 1960 and 1970. The first “high-level programming language” was <a href="https://en.wikipedia.org/wiki/Plankalk%C3%BCl" rel="nofollow ugc">Plankalkül</a> from around 1945. Even earlier, the roots of programming go back to Ada Lovelace, yes, a woman, in the year 1883.<br />
Nowadays, we even have low code or no-code ways to speak with machines. Mostly those use visual environments to drag together code components or nodes and are called visual programming languages.<br />
So, in the past, we needed to write code in a language, following grammar. Now, we could also drag together many nodes and hope for an effect, while playing with parameters. At some point of customization those visual environments also allow to add code manually. But it's not essentially necessary to write the code yourself anymore.<br />
DALL·E and other neural networks allow access to Artificial intelligence as a source of processing your thoughts via a text prompt. If you want, you could say that you tell the machine, what it should program with your natural language. Your chosen words define the quality of the output you are aiming for.<br />
Natural language and natural interfaces<br />
The natural language is the way we as humans speak with each other. To some extent even our voice-interfaces around us are now able to communicate with us in a way that feels almost natural. In this case our command prompt mostly starts with “Alexa”, “Siri” or “Hey Google” to initialize the dialogue. Natural language processing takes the task of converting speech into words in the background to take the right commands.<br />
In 2013, there was a trend to use natural language forms in interfaces to make form usage feel more natural. Forms were integrated into sentences rather than presented as a separate option. Stats are more usable in natural language, but confusing numbers make it difficult to understand the impact of improvements.<br />
The tendency to make interfaces more natural or more human, so we could use our natural senses to interact, is present. But the evolution of the interfaces also depends on technological developments.<br />
<a href="https://vimeo.com/420773167?embedded=true&amp;source=vimeo_logo&amp;owner=53957807" rel="nofollow ugc"></a>Venous materials from MIT lab – Watch example on Vimeo<br />
There are many interesting developments, especially if you investigate the MIT Media lab and the tangible media group. They strive for seamlessly coupling the worlds of bits and atoms by giving dynamic physical form to digital information and computation. <a href="https://www.media.mit.edu/projects/prototyping-interactive-fluidic-mechanisms/overview/" rel="nofollow ugc">Research on venous materials and human-flora interaction provides insights into our future.</a><br />
One of the easiest ways to interact now is our language, expressed by voice interfaces. When using Neuralink or similar chips inside our bodies, it's likely that we'll need to use interfaces with a well-structured language initially.<br />
If you are looking for more scientific terms on how the language is being used for generating visual results, look for <a href="https://www.google.com/search?q=text-to-image+synthesis&amp;rlz=1C5CHFA_enDE985DE985&amp;oq=text-to-image+synthesis&amp;aqs=chrome..69i57j0i19j0i19i22i30l5j69i61.1431j0j1&amp;sourceid=chrome&amp;ie=UTF-8" rel="nofollow ugc">text-to-image synthesis</a>.<br />
GANs produce visual effects resembling the WebGL example on alteredqualia.com that utilizes DeepWarp, a method commonly employed for gaze movement.</p>
<p>CLIP, GANs or AI driven Art and approaches of deep learning.<br />
Natural language processing is also an important tool when it comes to playing with so called CLIPs. CLIP is “Contrast Language-Image Pre-Training”, a neural network trained on a variety of text-image pairs. CLIP got introduced around January 2021.<br />
Before getting into CLIP, a short excursion on learning. Deep learning is part of machine learning and helps machines to learn. The machine is supposed to increase its own skills without the help of humans. On contrast to machine learning, where humans interfere in the learning process, in deep learning humans take care of providing information for learning. To do so, we could also use our input prompt. The process of analysis is up to the machine itself.<br />
An example for machine learning would be connecting words and images by labelling the images with the appropriate word. A typical combination for this cause would be <a href="https://www.image-net.org/" rel="nofollow ugc">ImageNet</a> and <a href="https://wordnet.princeton.edu/" rel="nofollow ugc">WordNet</a>.<br />
CLIP is a neural network model that is already trained on 400,000,000 (image, text) pairs. So, this means that this is the data that is used in training the CLIP Model. You can input an image to the model and get the most probable caption or summary of it. Alternatively, you can use natural language as an input. The true magic comes when you see examples. <a href="https://twitter.com/metasemantic" rel="nofollow ugc">Travis </a><a href="https://twitter.com/metasemantic" rel="nofollow ugc">Hoppe</a> shares some interesting applications in his Twitter account.<br />
To get a full “beginners” understanding of the CLIP model, I recommend this article: “<a href="https://www.kdnuggets.com/2021/03/beginners-guide-clip-model.html" rel="nofollow ugc">A beginner’s guide to the CLIP Model</a>”. You will find many interesting articles on <a href="http://kdnuggets.com" rel="nofollow ugc">kdnuggets.com</a> around artificial intelligence, computer vision, data science, machine learning, natural language processing or programming (Python).<br />
To try CLIP yourself, you could use <a href="https://compvis.github.io/taming-transformers/" rel="nofollow ugc">VQGAN</a> developed by the University of Heidelberg. It’s a GAN (a generative adversarial network), those are special types of neural networks consisting of “Generator” and “discriminator” networks. There is a GAN that generates images. The generator is like an artist and the discriminator is like a critic during training. When the generator can produce an image that the discriminator can't distinguish from a human-made one, we can use it for creating art.<br />
There are huge libraries of available GANs for very different purposes. Some of the lists I could recommend:<br />
<a href="https://github.com/hindupuravinash/the-gan-zoo" rel="nofollow ugc">The GAN Zoo</a><br />
<a href="https://github.com/nashory/gans-awesome-applications" rel="nofollow ugc">GANS awesome applications</a><br />
To play with the technology to learn, use <a href="https://www.artbreeder.com/" rel="nofollow ugc">artbreeder</a> and generate new images; Artbreeder is also based on GAN based image generation.</p>
<p><a href="https://www.artbreeder.com/" rel="nofollow ugc">Artbreeder</a>, Image generation based on GAN<br />
Exploring Applications of CLIP<br />
How to get started? I would always recommend playing with the technology. Most technology is too complex for an early deep dive, so start by playing. If you found your motivation to dig deeper, do it with all your might as you need a lot of patience to get a good understanding.<br />
So, let’s look at <a href="https://colab.research.google.com/github/justinjohn0306/VQGAN-CLIP/blob/main/VQGAN%2BCLIP(Updated).ipynb#scrollTo=JqFEaEChR20v" rel="nofollow ugc">VQGAN + CLIP</a> you could use directly yourself online. This is a collab notebook, a small development environment running Python programming language. Your task is to input some values... To get started, I would recommend to only change the “natural language prompt” of what you want to see. In my case I tested some results of different apps with the sentence “Children playing in the forest” (because my kids are attending a kindergarten in the wood while writing these lines).<br />
These are my first results using the CLIP linked above (to try yourself, open the <a href="https://colab.research.google.com/github/justinjohn0306/VQGAN-CLIP/blob/main/VQGAN%2BCLIP(Updated).ipynb" rel="nofollow ugc">link</a>, basically go through the page and always execute the play button to the left and wait till there’s a check icon next to it, to signal it’s loaded), only for chapter “Parameters” enter a text. In the end you could also create a video. The results here are early iterations. The program would continue till the “free” volume is used up.</p>
<p>First iterations with VQGAN + CLIP<br />
Number 100<br />
Number 150</p>
<p>from <a href="http://wombo.art" rel="nofollow ugc">wombo.art</a><br />
from <a href="https://creator.nightcafe.studio/" rel="nofollow ugc">nightcafe</a></p>
<p>To compare my result, I checked for other applications out there and their result. So there’s for example an application that is also available as an app; In <a href="http://wombo.art" rel="nofollow ugc">wombo.art</a> you could test directly. The result from wombo art had similar patterns as my test with YQGAN and looked like the picture on the at the top left.<br />
Then I used another popular web-service to compare: <a href="https://creator.nightcafe.studio/" rel="nofollow ugc">Nightcafe</a> you can see on the right.<br />
You see, when using a webservice, you end up getting one result. Using a <a href="https://colab.research.google.com/" rel="nofollow ugc">google colab</a>, you get many iterations and it's not as fast as the webservice or app. (Google Colab or Colaboratory is a way to execute Pyhton code in the browser).<br />
There are also other CLIPS on google colab you could use. The following picture is the (an early iterative) result of the combination <a href="https://colab.research.google.com/drive/1L8oL-vLJXVcRzCFbPwOoMkPKJ8-aYdPN#scrollTo=g7EDme5RYCrt" rel="nofollow ugc">VQGAN+CLIP (z+quantize method)</a>:</p>
<p>You could continue playing with many more version out there. Here you will find an overview of newer applications: <a href="https://softologyblog.wordpress.com/" rel="nofollow ugc">https://softologyblog.wordpress.com/</a><br />
You notice that the basic style follows one pattern and it’s obvious nightcafe also uses some way of VQGAN combination. To create magic, or what I would call magic based on a text prompt, there’s still some way to go.<br />
<strong>The magic I aim for: I tell the machine what I want to see, and they give me iterations of the thing I type in in. It mustn't be art or AI art</strong>.<br />
To create this magic, we look at DALL·E. <a href="https://openai.com/" rel="nofollow ugc">OpenAI</a> released DALL·E 2 in April 2022. They show the real magic of the future. By the way, I wouldn't call the results art. Mostly we generate graphics.<br />
DALL·E 2 screenshot, see more on their <a href="https://openai.com/dall-e-2/" rel="nofollow ugc">website</a><br />
A short introduction to DALL·E, DALL·E 2 and DALL·E mini<br />
<em>“DALL·E 2 is an artificial intelligence model that takes a text prompt and/or existing image as an input and generates a new image as an output. DALL·E 2 was developed by researchers at OpenAI to understand the capabilities and broader implications of multimodal generative models. To help us and others better understand how image generation models can be used and misused.”</em> (From their own description)<br />
To make it shot, look at <a href="https://openai.com/dall-e-2/" rel="nofollow ugc">their website</a> and get a feeling of the beauty that could be generated with DALL·E 2 or watch the <a href="https://www.youtube.com/watch?v=qTgPSKKjfVg&amp;t=93s" rel="nofollow ugc">short video</a> on youtube:<br />
<a href="https://www.youtube.com/watch?v=qTgPSKKjfVg&amp;t=93s" rel="nofollow ugc"></a><br />
Unfortunately, DALL·E 2 is not easily available to test yet. But you could sign up on their waiting list.<br />
According to OpenAI DALL·E 2 isn't trained on faces, to avoid the risk of deep fakes. The danger is already there. Finally it the time to doubt any images being shown to us!<br />
On twitter try a search for <a href="https://twitter.com/hashtag/Dalle?src=hashtag_click" rel="nofollow ugc">#dalle</a> or <a href="https://twitter.com/hashtag/dalle2?src=hashtag_click" rel="nofollow ugc">#</a><a href="https://twitter.com/hashtag/dalle2?src=hashtag_click" rel="nofollow ugc">dalle2</a> and you find many inspiring pictures. Going through the magical results, it feels as if image searches could get a relic of the past soon or will only be needed for special causes. Image stock market could be AI generated in my point of view. Here are some randomly selected shots from some twitter members (all linked):<br />
<a href="https://twitter.com/hardmaru/status/1524303013766373377" rel="nofollow ugc"></a><br />
<a href="https://twitter.com/djbaskin_images/status/1523816614713470976" rel="nofollow ugc"></a><br />
<a href="https://twitter.com/hardmaru/status/1523664409838510082?ref_src=twsrc%5Etfw%7Ctwcamp%5Etweetembed%7Ctwterm%5E1523664409838510082%7Ctwgr%5E%7Ctwcon%5Es1_&amp;ref_url=https%3A%2F%2Fpublish.twitter.com%2F%3Fquery%3Dhttps3A2F2Ftwitter.com2Fhardmaru2Fstatus2F1523664409838510082widget%3DTweet" rel="nofollow ugc"></a><br />
Yes, it’s a pity, we all can’t directly play with this fascinating technology!<br />
Luckily I found <a href="https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy" rel="nofollow ugc"><strong>DALL·E mini</strong></a><strong>.</strong> And that's your friend to play! By the way, I also created my post image with Dall-e mini. I asked for "AI Art blue and yellow". One of the results could be found on the overview site of my <a href="https://manfred-feiger.com/writing/" rel="nofollow ugc">writings</a>.<br />
The description of <a href="https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy" rel="nofollow ugc"><strong>DALL·E mini</strong></a>: <em>“OpenAI had the first impressive model for generating images through DALL·E. DALL·E mini is an attempt at reproducing those results with an open-source model.”</em><br />
They also have DALL·E mega, a large version of the mini, that is being trained. To see the fascinating world of training, look into the <a href="https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-Mega-Training--VmlldzoxODMxMDI2" rel="nofollow ugc">training space</a>.<br />
And finally, here’s your playground to play with the DALL-E mini: <a href="https://huggingface.co/spaces/dalle-mini/dalle-mini" rel="nofollow ugc">https://huggingface.co/spaces/dalle-mini/dalle-mini</a><br />
That’s my result on "playing children in the forest"; quite an improvement in a very short time. If you play more with <a href="https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy" rel="nofollow ugc"><strong>DALL·E mini</strong></a> you will notice, that faces still are really not trained much. But I already love it!</p>
<p>To feel the new power of words and to discover some magic along using words, go out and play for yourself!<br />
By the way, there's another tool on the web, giving you similar results. It's <a href="https://neuralblender.com/" rel="nofollow ugc">neuralblender</a>. Here's my kids results from them:</p>
<p>If you want to get inspired by other AI Artists, look at the work of <a href="https://manfred-feiger.com/create-ai-art-like-refik-anadol/" rel="nofollow ugc">Refik Anadol</a>, his process is the application of a lot of machine learning to a real world of art. As I said, in my point of view, the term of AI Art is mostly wrong, looking at the output of DALL·E or other CLIP applications, as the outputs are not expressive. In my point of view, we speak about AI graphics.<br />
Maybe you could guess, what my prompt in this case was (created with DALL·E Mini):</p>
<p><strong>Have fun playing!</strong><br />
And guess what... My challenge is to build websites via a text prompt. Currently I am looking for available sources to have a good segmentation (Structural attributes, Style attributes). Let's see when I will share my results or findings.</p>
<p>Some more tipps on playing with <a href="https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-mini-Generate-images-from-any-text-prompt--VmlldzoyMDE4NDAy" rel="nofollow ugc">DALL·E mini</a>... use objects. To avoid deep fakes, faces won't work. Play around with the text prompt. You could try with the attribute, in the style of... or try the <a href="https://docs.google.com/spreadsheets/d/1y7nAbmR4FREi6npB1u-Bo3GFdwdOPYJc617rBOxIRHY/edit#gid=0" rel="nofollow ugc">DrawBench</a> prompts being used for Google Imagen first.<br />
Dalle Mini: a rocket in the style of Matsuo Basho</p>
<p>Google Imagen<br />
While publishing this article on Mai 24th, Google came up with a competitor to OpenAI DALL·E 2. The market is on fire! And Google-<a href="https://imagen.research.google/" rel="nofollow ugc">Imagen</a> impresses the same way, DALL·E 2 did.<br />
Be excited to check out the first tweets on the topic <a href="https://twitter.com/hashtag/imagen?src=hashtag_click" rel="nofollow ugc">#imagen</a>.<br />
And we will all be happy to play with the first, real demos!<br />
Screenshot from <a href="https://imagen.research.google/" rel="nofollow ugc">Imagen</a> website</p>
<p>Recommended further readings</p>
<p><a href="https://mirror.xyz/herndondryhurst.eth" rel="nofollow ugc">Infinite Images and the latent camera</a><br />
<a href="https://arxiv.org/abs/2204.06125" rel="nofollow ugc">Hierarchical Text-Conditional Image Generation with CLIP Latents</a><br />
<a href="https://openai.com/dall-e-2/" rel="nofollow ugc">DALL·E 2</a><br />
<a href="https://imagen.research.google/" rel="nofollow ugc">Imagen by Google</a><br />
<a href="https://nvlabs.github.io/stylegan3/" rel="nofollow ugc">StyleGan 3</a><br />
Explore my newer articles around the topics:<a href="https://manfred-feiger.com/how-ai-assistants-help-you-with-text-and-art/" rel="ugc">How AI Assistants Can Help You with Text and Art</a><a href="https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/" rel="ugc">Generative Art – how to get started with AI Art</a><a href="https://manfred-feiger.com/midjourney-and-dall-e-2-first-hands-on-impressions/" rel="ugc">Midjourney and DALL-E 2 – first hands on impressions</a><a href="https://manfred-feiger.com/how-stable-diffusion-celebrates-new-forms-of-creativity/" rel="ugc">Stable Diffusion celebrates new forms of creativity</a></p>
<p>FAQs<br />
What are AI GraphicsVisual images or designs on some surface, such as a wall, canvas, screen, paper, or stone, to inform, illustrate, or entertain being generated with the help of Artificial Intelligence.<br />
<a href="https://manfred-feiger.com/create-ai-art-like-refik-anadol/" rel="ugc">Deconstructing AI Art masterpieces: Create Art like Refik Anadol</a>What If you were asked to create art like Refik Anadol? Or to produce some "AI art"?<br />
In case you<br />
<a href="https://manfred-feiger.com/midjourney-and-dall-e-2-first-hands-on-impressions/" rel="ugc">Midjourney and DALL-E 2 – first hands on impressions</a>Maybe you remember my article speaking computer language – welcome AI art, speaking about the new role of t<br />
<a href="https://manfred-feiger.com/generative-art-how-to-get-started-with-ai-art/" rel="ugc">Generative Art – how to get started with AI Art</a>The world of art is a diverse and beautiful place. The more we explore new forms of art, the more we are<br />
<a href="https://manfred-feiger.com/how-ai-assistants-help-you-with-text-and-art/" rel="ugc">How AI Assistants Can Help You with Text and Art</a>Have you ever wished you had a personal assistant to help you with your writing or with your artworks?<br />
AI</p>
]]></content:encoded>
		
			</item>
	</channel>
</rss>
