Tensor Art AI

 

I recently came across yet another AI imaging app - there seems to be a new one every week - named Tensor Art.  This one, at least on first impression, seems better than most I've tried and somewhat resembles Leonardo AI, though Tensor appears to offer even more features for fine tuning images including inpainting, the ability to add Loras (a premium feature on Leonardo) and specify seeds, a choice of adetailer models and sampling methods, a hires fix that not only offers a choice of sizes but also of the number of steps to achieve that size.  And Tensor, of course, also offers the more commonly found features such as a choice of aspect ratios and image-to-image prompts.  For a free app (a set number of credits are assigned daily with each image generation costing a single credit) it's really very impressive.  There isn't even the need to set up an account - one simply logs in using one's email address,

Tensor's image quality, if not quite up to Midjourney standards, is still very good,  Below are several images generated using one of the app's platform models.

For those interested in taking the app for a test run there's a useful YouTube tutorial by Making the Photo that concisely explains all Tensor Art's features in less than 15 minutes.

Leonardo AI Now Offers Loras

Leonardo AI, an excellent AI imaging app. generously offers a free tier of service to its users in the form of 150 credits a day with which to generate images.  The app is, however, a commercial venture and in order to increase its revenue stream has hit on the idea of adding new "premium" features, seemingly one every few weeks, that are only available to paying subscribers.  I can't really fault Leonardo for doing this since its subscription rates are reasonable enough (though they certainly have to be to remain competitive with Midjourney's) and the premium features themselves, such as Alchemy and PhotoReal, are generally very useful.

Leonardo's latest new paid feature is Elements, its own brand of LoRa (Low Rank Adaption) models that have the ability to greatly enhance the appearance of generated images, either on their own or in conjunction with the app's platform models.  Leonardo describes the new feature as follows:

"We are thrilled to announce that Leonardo Elements; our proprietary custom implementation of LoRA (Low-Rank Adaption) models are now available to use on the platform.

Taking advantage of Elements is easy; in order to get started hit the new ‘Add Elements‘ button under the prompt generation input:

Select from Elements you would like to use for your generation. You can select up to four Elements to combine, or use one on it’s own. Find and experiment with different styles, find your favorites - and go wild!

Once your selection is complete, you can adjust the weights for each Element to fine-tune your results. Adjust weights to get the perfect balance for your generation."

In addition to this introduction, there is a useful if overly long YouTube tutorial by Wade McMaster that tells everything there is to know about Elements and then some.  The video would actually have been better at half its almost 10 minute length.  As it is, there are a few too many samples shown and the accompanying commentary would have benefited by being more concise.

Leonardo has come a long way in only the past few months and now is fairly comparable to Midjourney in the quality of the images its generates and the features it provides.  It's the one app I feel makes it worth upgrading to a paid subscription, both for the aforementioned quality of its images and for its ease of use,  and I will probably do so sometime in the near future when I have more time.

Nikon Announces 135mm Plena Z-Mount Lens

Nikon has just announced the release of its 135mm, f1.8 full frame S Plena Z-mount lens which it has been teasing for the past several weeks.  In case you're wondering, the "S" in Nikon's S line apparently stands for "Supreme Optics and Design," and judging from its specs and the sample photos on Nikon's website this quality optic certainly appears to merit that designation.

135mm has always been every camera manufacturer's most commonly used focal length for portrait (i.e., moderate length telephoto) lenses, but unless the sample photos lie there's nothing common about this optic that boasts the following features:

• Aspherical design;
• Extra-low dispersion (ED) glass;
• Short-wavelength refractive (SR) glass for precise chromatic aberration compensation;
• Meso amorphous coating;
• Arneo coating;
• Stepping motor;
• Auto-priority manual mode;
• Internal focusing that prevents the lens changing in size during auto-focusing;
• Electromagnetic diaphragm mechanism

Impressive as these specs are, it's the quality of the sample photos that is truly outstanding.  The bokeh produced by the Plena is nothing short of amazing, and the color rendition is nearly perfect.

At a list price of $2,496.95 the Plena is certainly not for every photographer, but if it lives up to its advertising I'm certain it will become a must-have for every serious portrait and wedding professional as well as those bokeh enthusiasts willing to pay a premium to achieve effects such as those shown in the sample photos.

Adobe Firefly

As I'm always interested in experimenting with new AI imaging apps, I recently turned to Firefly after having received an email from Adobe last week inviting me to try out the newest version.  I had actually taken a look at the app when it had first been announced by Adobe several months ago and had not been particularly impressed by what I had seen.  (See my March 23 post.)  The text-to-image results I obtained today, though, were far superior to what I had seen earlier in that they were much more photo realistic.

In addition to text-to-imaging capabilities, Firefly has several intriguing new features - generative fill, text effects, generative recolor, sketch-to-image, and 3D-to-image - and has several more in the works, including personalized results, text-to-vector, text-to pattern, text-to-brush, extend image, and text-to-template.  I haven't yet tried any of these myself but they certainly appear promising.

As far as the text-to-imaging results I obtained on my latest try, and with which I was fairly well pleased, I am posting a few samples below.  (Note that all the human figures generated by Firefly were minority members even though I had not requested this.)  As a prompt, I used text generated by ChatGPT originally intended for use in another AI imaging app.  The prompt reads as follows:

"In the dystopian metropolis of NeoCity, a weathered and weary replicant, known only as X42, finds solace in recounting its extraordinary experiences to a disenchanted group of rebels. Its voice resonates with both longing and resignation as it speaks: 'I've seen things you people wouldn't believe. Attack ships on fire off the shoulder of Orion. I watched C-beams glitter in the dark near the Tannhäuser Gate. All those moments will be lost in time, like tears in rain.'"

Whatever its literary merits, the prompt successfully gave me the cyberpunk vibe which I was seeking.

As I was downloading the images generated by Firefly, I noticed that for each a label appeared that read: "content credentials applied."  I then clicked the Learn More button to discover exactly what Adobe meant by "content credentials."