This AI-powered art app lets you paint pictures with words – TechCrunch

0
110

The power to bend the reality of AI has become more and more busy over the last decade. There is a 3D diorama that uses computer vision. Trend setting style transfer; viral photorealistic selfie tuning, selfie retouch, face swap, and — ofc — deepfake.And a lot of frivolous (and Hilarious) Insert a selfie filter (ohhai’Disneyfying’ cartoon lens!) And have fun.

AI-powered visual remixes can get a lot of attention over and over again. When the novelty of the effects generated by AI disappears, it becomes difficult to “keep an eye on”. (Selfie retouching apps don’t have that problem. There is a constant demand for machine learning as a reality. Enhancer.. )

Most notable in the development of AI-enabled synthetic media during this period is how fast these visual effects have improved with the help of more powerful mobile processing hardware than ever before.

The time to wait for the finished result is essentially instant. A game changer for commercializing (and potentially monetizing) the creativity and power of neural networks and GANs (Generative Adversarial Networks). alias, Machine learning frameworks perform retouching, reframing, or actually generated modeling and jump from human prompts for inspiration.

Also, most of the app-based visual remixes over the last decade have focused on retouching / styling / enhancing and pure AI-based image generation, but that’s changing too.

Wombo, a Canadian startup named after an AI-enabled lip-sync video app, recently used AI to create original “artwork” based on text prompts, another called Dream (iOS and Android). App has been released.

That’s great, splendid Simple: Briefly explain what you want to paint. For example, choose a style from the options offered, such as “Awesome Tree” or “Worst Sandwich Ever” (mysterious, baroque, fantasy art, steampunk, etc.), or press “No Style”; Create.

Then, literally in a few seconds — I counted <20 — the app will display the completed “artwork”.

You won’t even get bored with these few seconds of creation because you can get a glimpse of the AI ​​in action. The app shows a rapid evolution of modeling, from starter marks to some inhumane fast additions to flesh out the canvas, virtually breathtaking to reach another finished configuration. ..

Some of these generated artworks are a bit impressive. Some … not so many.

However, of course, the two prompts do not produce the same image. So you can keep requesting new images from the same prompt until you like it.

In short, Christmas card artists and pulp fiction illustrators can probably retire now.

Everyone is now an “artist”.

That said, the actual artist should do it Don’t worry. In particular, art created by the human brain and body becomes more valuable only when the world is flooded with “machine art.” (May all NFT mint dilute the meaning of the phrase “digital art” …)

The quality of the “art” of the Dream app will definitely fluctuate. Longer and more complex prompts seem to confuse it. Therefore, the quality of the output depends on what you request to draw.

That “style” tends to be more abstract and distorted than a particular exact one, if it can be thought of as having a single inclusive style in the midst of so many pastiches. I have. Therefore, portrait requests are not rendered photorealistically. And it’s usually more comfortable to draw something fantastic than it really is. (The “Madonna and the Child” prompt, for example, provided work closer to the failure of the restoration of the infamous Spanish church than Crypto Botticelli.)

However, the speed of production is impressive. Horrible slash.

As soon as you see fresh artwork, the app doesn’t waste time trying to sell it. An option to “Buy Printed Matter” will pop up. It links to a web shop and looks like a clever way to turn visual tricks into real revenue. (We offer “Custom Wombo Dream Print”[s]Starts at $ 20 for matte posters and $ 45 for framed prints. )

If a startup can turn about 20 seconds of processing into revenue of $ 20 or more, it’s a great little money pipeline.

By the end of last month Over 10 million images It was already generated by the user. (Although the number of downloads of the Google Play app has already exceeded 1 million times, A month or so After release)

However, most people have a finite wall space to hang all kinds of art. Not to mention the images produced by ignorant machines. Therefore, most of these random pieces remain tightly virtualized. (Although “AI Art” may be the perfect NFT feed …)

Where “AI art” lies in fashion / cultural values ​​is certainly an interesting question to ponder.

Indeed, it’s better than clip art and stock photography. Also, the output of the Dream app can be more interesting than the average “art” print you can buy at IKEA. However, the results can be quite weird, derivative, nappy, or just weird.

And, well, is it art? Or is it just a visual output of the mathematical process? An abstraction of human creative skills that can’t translate real emotions, identities or soul sensations because the code doesn’t have those things? I’m just doing what I was told.

And do you really want to put a coded abstraction on your wall?

So maybe?if In particular aesthetic. But, well, is it art or just wallpaper? Maybe Wombo should sell a roll of Dream AI wallpaper or a printed mouse mat and T-shirt (March) instead of an “art” print …

There is a lot to ponder.

There are some obvious things. The art generated by AI is a lot of fun to play with. It’s a kind of visual catnip. Imagination toys.

Also, definitely stay here. And AI models continue to be “better”, depending on what “better” means around subjective subjects such as art. (Perhaps the generative art model brings more successful results by completely engaging the user in the creative process. It provides tools for customizing and manipulating the output of the machine, one imagines. Allows you to make adjustments (fine adjustments) until you get closer to something, or otherwise felt It’s more personally unique and meaningful. Or, in other words, a more hybrid creation process can produce a more powerful and inspiring artistic output. )

There are also these artistic AI scores, each of which produces different “flavors” and “characters” of visual output derived from training data. Alternatively, you can use different “style” art AIs, if desired. (But perhaps the “specialty” is closer to a coded mark.)

There are many other GAN-based image generation AI tools — and I Big fan Pixray’s system (pixel art output is especially cute) is much slower, but Wombo seems to have been the fastest punch to apply and monetize this technology.

The next decade of machine learning that bends reality will be a significant journey.

LEAVE A REPLY

Please enter your comment!
Please enter your name here