Creating AI Images from Brain Waves

In my recent sci fi/fantasy novel And What If, I described an instrument capable of generating AI images from a subject's dreams.  Although the idea was of course speculative, I had based it on actual research done in Finland and in Japan that had succeeded in transcribing thoughts into images by measuring the flow of blood in the brain.  It seems now that exciting new research has taken the idea one step further.  According to a paper published in November entitled Seeing Beyond the Brain that is the subject of an NBC news article a team of researchers led by Zijiao Chen that includes participants from the National University of Singapore, the Chinese University of Hong Kong, and Stanford University "have shown they can decode human brain scans to tell what a person is picturing in their mind."  The NBC article from which this quote is taken then goes on to describe the process in detail.

Although no one yet, as far as I know, has succeeded in generating images from a subject's dreams as I hypothesized in my novel, it will obviously be only a matter of time until such a breakthrough occurs.  No doubt someone reading my novel ten years from now will find it laughably outdated.

DPReview to Close

According to an announcement on its site DPReview will cease operations as of April 10.  The site will be sorely missed, at least by me, as I found its product reviews to be among the most reliable available.  Few other sites perform the in-depth testing DPReview routinely provided.  Although I sometimes disagreed with its conclusions, they were always worth serious consideration.  The site was also a good source for industry news.

I had not realized, prior to reading the closure announcement, that DPReview had actually been wholly owned by Amazon which has been engaging recently in a wave of cost-cutting moves.  While I appreciate that Amazon needs to turn a profit, surely it could have found a better way to save a few dollars than by terminating a valuable service.  After all, it is not as if the company were in any sort of financial difficulty.

Adobe Announces Firefly

I received an email yesterday from Adobe Stock announcing the beta version of Firefly, it's new app for authoring AI-generated content.  Glancing at the images shown at the above link, I don't really see much here. other than text effects and vector recolors, that isn't already offered in other more established AI-generating apps.  In fact, the text-to-image examples shown seem a bit cartoonish, especially when compared with the lifelike renderings that can be achieved with Midjourney.  Most noticeably, there is no text-to-video option which is what I think everyone is most interested in right now after the recent announcement of Gen-2 by RunwayML (though of course for stock image purposes Adobe would naturally be more interested in stills).  Nevertheless, I requested access from Adobe and will experiment with the beta version once my request has been granted; I will then post the results here.

Warning on Adorama Camera

 

I just wanted to give everyone a warning on doing business with Adorama Camera in NYC where I've just had a very unpleasant shopping experience.

Briefly, I went into the Adorama store on West 18th Street on Monday to purchase a Rode tripod that the company's website had listed as "in stock."  When I arrived at the store the salesman told me that although the item was in stock, it was not in stock at the store and that there would be a wait to either have it delivered or picked up.  Since I was already on the premises I decided to purchase it anyway and then pick it up at the store at a later date since the salesman assured me it would be there either on Wednesday or, "at the latest," on Friday.

Today, Tuesday, out of simple curiosity, I checked my Adorama account online to see the status or my order.  To my surprise it had been canceled with no reason given.  After first chatting online with a rep who ultimately told me she couldn't help me and that I would have to call the store I wasted another half hour with a customer service rep who finally told me I would have to speak directly to the salesman with whom I'd placed the order to find out exactly why it had been canceled but that that salesman wasn't available just then because he was waiting on other customers.  When the salesman finally called me back some time later it turned out he had stupidly canceled the order in error.

The bottom line was that Adorama never contacted me to tell me that my order had been canceled.  If I hadn't bothered to check my account I would still be waiting to have my order filled.  It simply isn't fair to treat customers in such a cavalier manner and I've made up my mind never to do business with Adorama again.  I'm only posting this now to give fair warning to my readers.  Caveat emptor.

Runway Announces New Text-to-Video App

Those who've been following this blog know, I've been working intensively in recent months in authoring AI-generated imagery through the use of such apps as Stable Diffusion, DALL-E 2, and Lexica.  As a photographer, my greatest interest up to the present has quite naturally been in creating still images.  Now, however, just as I've begun working seriously with video for the first time in my career, an article in The Verge has announced that a start-up company called Runway is about to make available to the public a new text-to-video app called Gen-2, a follow-up to the far more primitive Gen-1.  Given the almost rabid interest shown it the text-to-image apps named above it was probably inevitable that something like this would soon appear.  It's not likely, though, that anyone is going to be making full length motion pictures with Gen-2 in the foreseeable future.  The examples I've seen are only 4 seconds long and even at that short length are too jumpy and lacking in anything like 4K resolution to be usable.  (As The Verge article notes, "...all we have to judge Gen-2 right now is a demo reel and a handful of clips [most of which were already being advertised as part of Gen 1.]). Nevertheless, it can safely be assumed that the specs will eventually improve to the point that viewers will be unable to differentiate AI-generated videos from those created with traditional media.

Runway's pricing plans seem quite reasonable, but until the app shows greater improvement I will most likely stick with the free basic plan that should be more than sufficient to enable me to experiment with creating short AI-generated videos of my own.

Synthography

 

I've recently come across a new term in the rapidly changing world of AI-generated art.  According to the brief definition in Wikipedia, synthography is "the method of generating digital media synthetically using machine learning."  Since this pretty much describes the process I've been using in such apps as Stable Diffusion, DALL-E 2, and Lexica in authoring AI-generated images through the use of text prompts, I've taken the username nyc-synthographer on Tumblr and nyc_synthographer on Instagram.  I've only joined both those sites recently and up to now have been content with posting examples of my traditional photography as I find my way around.  In the near future, however, I intend to use the two sites as forums for the publication of my AI-generated images.  Stay tuned.

US Copyright Office Issues Guidance on AI-Generated Imagery

 

The US Copyright Office today issued fresh guidance on copyrighting images derived from text prompts that have been entered into various AI imaging apps.  Since I'm not an attorney and therefore not qualified to render a legal opinion, I will let readers peruse the document for themselves and make what they can of it.  I for one find it extremely confusing.  The Office first seems to be denying the possibility of copyrighting such works ("If a work's traditional elements of authorship were produced by a machine the work lacks human authorship and the Office will not register it.")  That's clear enough, but the same document also holds that photographs are eligible for registration "so far as they are representative of original intellectual conceptions of the author" even though such images are obviously produced mechanically through the use of a camera.  If that were not confusing enough, the Office then appears to hedge its bets entirely by stating, "In other cases, however, a work containing AI-generated material will also contain sufficient human authorship to support a copyright claim."  To say then that copyrighting AI imagery is a grey area is a vast understatement.  Those seeking copyright registration of AI-generated images will have to use their own judgment in making such submissions and see how the Office responds.

How DALL-E 2 Creates Images

In an interview in Venture Beat, Aditya Ramesh, Open AI creator of DALL-E 2, was emphatic that the AI app no longer scrapes images from the web in order to generate its output.  He stated in no uncertain terms:

"Diffusion models start with a blurry approximation of what they're trying to generate, and then over many steps, progressively add details to it, like how an artist would start off with a rough sketch and then slowly flesh it out over time."

That sounds fine and I would never want to question Mr. Ramesh's word, but his description unfortunately does not quite explain the above image created with DALL-E 2 in response to a fairly straightforward text prompt that ran as follows:

"Late 19th century elevated subway line running past tenement buildings on Manhattan's Upper West Side as seen from the street below.  Gloomy cloudy sky overhead.  Digital art in the style of John Sloan."

As you can see, the image fits the prompt admirably.  My problem is that although I did not name any specific tenement building in my prompt, DALL-E 2 nevertheless came up with a remarkably accurate rendition of the Endicott Hotel on Manhattan's Columbus Avenue between 81st and 82nd Streets, a building I know well if only because it's located almost directly across the avenue from my own apartment building.  It simply doesn't seem possible in light of Mr. Ramesh's explanation that the Endicott's facade could be rendered so exactly, all the more so as an elevated subway line, the Ninth Avenue IRT, once did in fact run opposite the Endicott's third floor windows until it was finally torn down in the 1930's. 

All this came to mind because I'm in the process of drafting of vlogging entry on the Endicott in which I preface my video with the above shown still image.  It's a near perfect representation.

Open AI Releases Chat GPT-4

Open AI has announced it is releasing Chat GPT-4, the newest version of the online app that's taken the internet by storm with over 100 monthly active users making it the fastest growing consumer app in history.  While most articles reporting on this release, such as that on the ABC website, have focused on how much more intelligent than its predecessors the latest version is, what to me is most intriguing is that GPT-4 will be multimodal, meaning that among other things it will have the ability to generate video from text prompts.  It will also have the ability to analyze visual images and provide general information regarding them.  It's truly astonishing how quickly the app is improving in its capabilities and makes one wonder if the time is near when it will actually achieve singularity, the dream of every AI user.

Using AI Imagery to Create a Graphic Novel

I recently came across an article in Japan Today regarding the use of AI imagery (in this case generated by Midjourney) to create a full length manga, apparently the first ever to be published in Japan.  What most caught my attention was the claim by the author, who goes by the name of Rootport, that he had managed to complete the work, illustrations and all, in only six weeks.

Predictably, Rootport's accomplishment has already created a great deal of controversy in Japan where manga art is regarded not only as an art form but also as an important source of income for the large number of artists who work in the field.  However, at least one prominent artist, Madoka Kobayashi, seeing the possibilities offered by AI, has gone on record as saying, "I don't really see AI as a threat - rather I think it can be a great companion." 

At any rate, the Japan Today article gave me an idea for my own next project.  Shown above is the cover of my noir novel The Blue Hours that I published online several years ago and that is still currently available for purchase on both the Amazon and Barnes & Noble websites.  I've always thought this one of my more successful works of fiction and have now decided to recast the story, with certain necessary modifications, as a graphic novel with illustrations generated entirely by various AI imaging apps.  The idea is all the more agreeable as I recently completed a fully illustrated novel, And What If (see my Featured Post), that enabled me to appreciate the extent to which illustrations can enrich text even in a work of fiction.

I most likely won't begin work on my projected graphic novel until autumn as during the warmer months I spend as much time as possible outdoors working on my photography and video, but it will certainly give me something to look forward to when the weather once again turns cold.  And even in summer months I may find time to spend some rainy afternoons experimenting with AI imagery for possible use in the book.

Coloromo

When on Facebook recently I saw a link to a new AI imaging app called Coloromo that allowed one to upload an image on which the app would then at no charge generate a number of watermarked thumbnail-size variations.  The user was then given the opportunity to purchase larger non-watermarked copies of these same variations (though there was nothing to stop a user from cropping out the watermark and enlarging the resulting image in an app such as Topaz Labs' Gigapixel).

I gave Coloromo a try and got back, among others, two variations on my avatar.  I noticed immediately that the styles of all the variations I received were suspiciously similar to those generated by the AI ReMix filter in Topaz Labs' Studio 2.  I don't know if this is coincidence or it the creators of Coloromo reverse engineered the Topaz Labs' filter, but in any case I see no further need to use Coloromo when I can get the same results much more easily in Studio 2.

Photoville Returns

I just read an announcement online that Photoville will be returning to Brooklyn Bridge Park this summer from June 3 through June 18.  

I last attended the photo festival in September 2019, only a few months before everything in NYC was shut down by the pandemic.  I don't remember having viewed any works of genius while I was there but there were a number of interesting displays, certainly enough that to warrant a return visit in June.  I'm always happy anyway to have an excuse to visit Brooklyn, the borough where I was born but have spent so little of my life.  (The night of my birth was actually the only one I ever stayed there.  The next day my parents took me home form Wykoff Heights Hospital to Ridgewood in Queens where they were then living.)

The photos shown here were taken during my 2019 visit.

Problems with Corel Video Studio 2023

I had posted late last month that Corel had released Video Studio 2023 Ultimate, the most recent update to its popular video editing app.  A few days later I purchased the upgrade and tried to install it on my Windows 11 Acer Concept D7.  The installation failed after I received an error message indicating I had an app open that needed to be closed.  Unfortunately, I was unable to locate the app on Task Manager and now believe it to be part of Corel's own installer.  At any rate, I was unable to close it.  I then tried chatting with a support technician who told me to reinstall Video Studio using the Repair option.  When that did not solve the problem the tech asked me to send her a printscreen image of the error message.  The message must have been as unclear to her as it was to me because she immediately told me she would set up a ticket and abruptly ended the chat session.

Left to my own devices, I tried doing a  hard reboot of the computer and again reinstalled the app using the Repair option.  This time the installation finished and the app loaded without further problem.

I thought that would be the end of the matter, but when I attempted to open the app yesterday it first failed to load and on a second attempt loaded but immediately froze so that I had no alternative but to shut it down with Task Manager.  It was only on the third attempt that the app loaded successfully.

In the meantime, in response to the ticket that had been created, I received an email from Corel tech support containing links that when opened contained instructions so convoluted they were impossible to follow, at least for a non-tech person such as myself.

I don't know if anyone else has experienced similar problems, but I wanted to post a heads up so prospective purchasers could be made aware of the problem.  Caveat emptor.

Panasonic MFT Cameras to Adopt PDAF

Building on the success of its full frame S5 II's Phase Detecting Auto Focus (PDAF) system, Panasonic has signaled it will begin installing the same system in its future MFT cameras as well.  An article in Photo Rumors quotes an interview PhotoTrend held with Yosuke Yamane, director of Panasonic's Imaging Business division, who was quite emphatic on this point.

"We will consider adding phase detection AF to our cameras depending on the model characteristics, not only for full-frame cameras, but also for Micro Four Thirds.  And on this last point, I invite you to stay tuned for our next announcements."

The main advantage of PDAF, in which the camera creates two copies of an image and then adjusts the lens elements until the two images merge, i.e., are in phase, is speed.  While this is not so important to photographers such as myself who shoot mainly stationary subjects, it is of great use to those who shoot moving subjects, in sports photography for example.  While I myself have never had any great problem with my Lumix G9's focusing ability, it has in the past received a large number of complaints from other photographers for its lack of speed, and this has proven a great disadvantage to Panasonic in maintaining an edge over its competitors. Correcting this flaw is obviously of extreme importance to Panasonic in its future marketing of MFT cameras. 

The other point to be taken away from Mr. Yamane's interview is that if Panasonic is willing to go to all the trouble of introducing a new focusing system in its MFT cameras then it is not likely to abandon that format anytime soon, an important point at a time when the industry as a whole appears to be moving away from MFT as witnessed by Sigma's recent decision to stop developing MFT lenses due to decreasing demand.

Midjourney's Attempt to Block Porn

One of the most persistent problems in AI imaging is its use creating highly realistic porn.  The advantage it offers to pornographers in obvious - they can have as many actors as they choose at no cost to themselves performing acts that are only limited by the imagination.

Now Midjourney, known for creating the most lifelike representations of human figures, is taking steps, however tentative, to correct the problem.  According to a report in the MIT Technology Review,  Midjourney is now banning all words related to the human reproductive system even though many of these words - such as cervix, placenta, and urethra - are not likely to be included in the vocabulary of most pornographers.  According to David Holz, Midjourney's founder, the ban is a "stopgap measure to prevent people from generating shocking or gory content while the company 'improves things on the AI side.'"  How successful this attempt will prove remains to be seen.

Meike Announces New Macro Lens

Those looking to shoot macro photography on a budget might do well to look into the new 60mm f2.8 entry from Meike that's designed for APS-C cameras with a range of different mounts available.  As  noted on the manufacturer's website, the lens has a magnification ratio of 1:1 and a minimum focusing distance of 0.175m.  The equivalent focal length will depend on the camera used.

I don't own any Meike lenses myself, but I've noticed they usually get good, if not stellar, reviews.  And at $189.99 the price is certainly right.