Four takeaways from our Jan. 19

Visual 1st Generative AI Spotlight

1. Generative AI for visuals creation: we’re not even at 1.0 yet

New technologies and newly addressed use cases are still emerging; we haven’t even passed the peak of Gartner’s Hype Cycle yet

• The underlying AI technologies available for photo and video app developers are still rapidly developing, with new releases coming out in rapid succession. 

• Many innovative photo & video app developers are already building apps that address a rapidly expanding variety of use cases. For instance, our 8 Visual 1st Generative AI Spotlight presenters showed solutions focused on consumer creative expression, SMB e-commerce, big brand e-commerce, business stock photos & videos personalization, and photo print product creation.

• The solutions for creating visuals through generative AI have not even passed the peak of Gartner’s Hype Cycle yet. Therefore yes, we’ll soon see disappointment and naysaying – before the dust settles and the true scope of opportunities for generative AI-based apps for photo and/or video creation comes into focus. 

Next Spotlight, March 15:

Photo Print App Innovation

Live demos and discussions with by 8 innovative photo print app developers

Attendees: Startup founders and corporate executives from the global photo ecosystem

Presenters to date:

Thomas Beguin, Co-CEO, IXXI

Kirandeep Kaur, Business Manager Mobile Commerce, ip.labs

Axel Lilienblum, Founder & CEO, Grow

Malhar Patil, Co-Founder, Zingcam

Buy your ticket now! More info.

$19 Early Bird Tickets expire February 28.

Sponsors to date:

2. Prompt-based image creation ain’t easy

To the rescue: Supplemental UI, large quantities, additional image correction/enhancement features, pre-made visuals, pragmatic video enhancement

The awe-inspiring examples of generative AI imagery that we’ve all seen might lead us to believe that prompt-based image creation is much easier than it actually is. In fact, as we heard repeatedly during our Generative AI Spotlight: it’s pretty darn hard to create great visuals through text prompts. 

Our presenters mitigate this challenge in several ways:

• Provide sliders or other UI elements that instruct the generative AI engines as to what type of visuals needs to be created, rather than the user needing to specify this all through text input.

• Enable users to generate a large number of different images from the same prompts, while setting the expectation that not all will be useful. In other words: acknowledge it’s a bit of hit or miss and leave it up to the user to select the image(s) that meet their expectations.

• Provide additional (AI) image correction functionality so that the image can be automatically corrected if needed, or make it easy for the user to do so manually. For instance, several of our Generative AI Spotlight presenters deploy automatic AI upscaling technologies to deliver the required resolution for their generative AI-created images.

• Give the user a choice of either acquiring pre-made generative AI stock images created in-house by prompt experts, or enable the user to create their own imagery through prompts. 

• Focus on those areas of video enhancement or customization that are feasible with today’s technology, such as altering the spoken words in video in combination with lip-synching, or transcribing spoken words into text and leverage this text to generate images for B-roll inserts.

3. The sweet spot: customization at scale

Generative AI’s sweet spot is the ability to create custom visuals at a fraction of today’s scale and cost

• Think of big brands running globally localized campaigns with custom imagery portraying a range of consumer persona.

• Think of SMBs who can’t afford professional product shoots.

• Think of stock photography buyers who need their photos to look “less stocky” or can’t find the exact image they had in mind 

4. Addressing image rights concerns – a must for big brands. And for the market to take off.

Concerns about the legal status of visuals created through generative AI will hamper adoption by big brands and, without that adoption, the generative AI market will only reach a fraction of its potential

If the concerns of creative professionals regarding the use of their copyrighted images in AI training sets were not already enough reason to develop ironclad practices and/or legislation, big brands’ image rights concerns will definitely tip the scale: without clarity on the legitimacy of using generative AI-created visuals, brands won’t adopt this new technology.

And a few more things...

Snapbar. Self-guided professional headshots. Visual 1st alumni Snapbar, which started as a an event photobooth solution and pivoted to online photosharing during COVID, has released a beta version of Snapbar Studio, a self-guided and admin-controlled solution for businesses to enable their employees to create their own headshots for HR.

Xpozer & Imaging Solutions AG. Partnership & investment. Xpozer, developer of wall décor mounting systems, and Imaging Solutions, manufacturer specializing in image enhancement software and modular systems for the automated production of on-demand photo prints, layflat books and canvas framing, have engaged in a partnership that includes Imaging Solutions distributing Xpozer products. Imaging Solutions also acquired a major share in Xpozer.

WhatsApp. Images at original resolution. Word is that WhatsApp is going to enable its users to share photos in their original quality without any compression. Currently, Whatsapp offers three photo quality options to select from: “Auto,” “Best quality,” and “Data saver.” However, even at the “Best Quality,” the app still compresses the image to accommodate fast data transfer.

Vyrill. In-video Search API. Visual 1st alumni Vyrill, developer of video intelligence solutions, launched its Video API, which provides access to Vyrill video intelligence and video commerce functionality. The API gives customers the ability to manage, rate, search and leverage their video data while accessing Vyrill's comprehensive insights and engagement data.

Fotor. Launching generative AI feature. Another photo app vendor jumping on the bandwagon of generative AI. Besides straight text-to-image creation, Fotor’s AI image generator also enables users to select a large variety of themes for their AI-created images, such as 3D, realistic, oil painting, cartoon, illustration, cyberpunk, Japanese manga, Ukiyo-e, new Art, psychedelic pop, concept art, photography, and other styles.

Taylor. Print on Demand. Graphic communications company, Taylor launched Taylor OnDemand, a new business united dedicated to developing new Print on Demand products and manufacturing capacity.

Lunchbox. Generative AI for food. Newly launched AI food photo generator, Lunchbox targets restaurants that need to create images of food for their social media pages, websites, and menus. Users have the option to select different backgrounds or styles of their imagery, such as a “polaroid,” “iPhone photo,” or “film camera.” Users can also generate their images as a “cartoon” or “painting” instead of a photo.

Shutterstock. Launching generative AI toolkit. A case of embracing one’s market disrupter: Customers of Shutterstock’s Creative Flow online design platform will now be able to create images based on text prompts, powered by OpenAI and Dall-E 2. The images are then “ready for licensing,” right after they’re made, according to Shutterstock. The service is part of a broader strategic partnership with OpenAI, which includes OpenAI using Shutterstock libraries to train and feed their algorithms. 

OpenAI. AI to detect AI. ChatGPT creator OpenAI released a free web-based tool designed to help educators and others figure out if a particular chunk of text was written by a human or a machine. Where’s the foolproof equivalent to do the same for generative AI-created visuals?

Watermark Remover.io. But on the other hand… AI-based Watermark Remover.io removes watermarks through just a few clicks, according to the Verge, who successfully used it on stock imagery from Shutterstock and Adobe Stock. 

Ethical AI Startups. Infographic. Apparently, we now also need AI to detect images that have had their watermarks removed through AI! – it’s a never-ending race between the bad guys and the solutions fighting them. Are there any good guys then? Reportedly, here are a bunch: check out the Ethical AI Startup Landscape infographic.

Best,

Hans Hartman