|
VideoGen. AI Videos Beyond Short Clips.
Today’s AI video tools can be entertaining and help cut commercial production costs, but most still generate isolated short clips. VideoGen aims to change that with VideoGen Model 1, a new model that turns text prompts and reference images into finished, multi-shot videos and short films up to 10 minutes long.
Unlike mainstream competitors such as Sora, Veo, and Kling — which typically produce 5- to 60-second clips that creators must manually stitch together — Model 1 acts as an end-to-end director and editor. It handles creative direction, screenplay planning, dialogue, sound effects, and music while maintaining visual consistency across scenes.
Check out the examples and read the company’s white paper.
Adobe MAX. Walking the AI Tightrope.
Adobe is walking a fine line: keeping pace with AI-first competitors without alienating its core Creative Cloud customers, who want to retain creative control and have legitimate concerns about being replaced by newcomers armed with AI tools.
The company’s latest balancing act was on display at Adobe MAX, where it announced new AI features while emphasizing that these tools support creative professionals rather than take control away from them. The announcements include:
-
Expanded AI agent support. Adobe Creative Agent is expanding across the company’s flagship Creative Cloud desktop apps to orchestrate multistep workflows, with local MCP integrations also on the way.
-
New Firefly features. Adobe expanded its Firefly generative AI platform with Firefly Design, Firefly Graph, and Firefly Builder, bringing raster, vector, motion, 3D, and no-code prototyping tools into one workspace.
-
New editing and enhancement tools. Additions across Adobe’s core apps include Relight and Generative Restyle in Photoshop, plus Color Mode and Generative Remove in Premiere. Following its acquisition of Topaz Labs, Adobe also introduced two AI image-enhancement models, Wonder 4 and Image HDR, for one-click upscaling, denoising, and dynamic range recovery.
-
Availability. Most of these Creative Cloud updates are available now. Several Firefly and AI Assistant tools are launching in beta, while local MCP integrations are scheduled to enter private beta in early 2027.
And beyond AI? Adobe’s best new photo features aren’t AI at all, according to DPReview. Spoiler alert: look to Adobe Camera Raw and Lightroom for those additions.
Aura. E-Paper Photo Frames Are Growing Up.
In the Amazon: The Photo Frame Gets Smart article in our last issue, I highlighted how Amazon’s new Alexa Photo Frame appeared to successfully combine AI curation, conversational photo search, smart side-by-side pairing, and voice notes. That’s meaningful innovation in a market that has traditionally evolved at a leisurely pace.
But intelligence isn’t the only thing changing in digital photo frames. Another important development is the growing use of e-paper displays. Rather than relying on illuminated pixels, e-paper uses electrical charges to rearrange ink particles, creating an image that looks more like a print than a screen.
Case in point: Aura has expanded its cord-free e-paper lineup with the 10-inch Aura Ink ($349), a smaller, sharper follow-up to its original 13.3-inch frame. The new model combines a nine-color E Ink Spectra 6 display with proprietary dithering and custom color tuning to mimic a printed photograph, with particular attention to natural-looking skin tones.
It also offers a higher pixel density of 200 PPI, an updated monochromatic design, and a motion-sensing front light that provides gentle illumination when needed. Battery life is rated at up to three months per charge.
Disney & Adobe. AI Partnership.
Disney is on a partnership spree. After announcing a deal with Shutterfly, which I covered here, and a short-lived deal with OpenAI, the company is now teaming up with Adobe.
Announced at Adobe MAX, the partnership aims to put Adobe’s current and future generative AI tools to work across Disney’s businesses, from theme park development and cruise lines to Walt Disney Studios and ESPN. The companies also plan to collaborate on “responsibly developed AI” to improve internal creative workflows and find new ways to engage sports fans.
The agreement builds on an existing relationship: Disney’s Imagineering R&D team already uses Adobe’s Firefly AI platform for park design.
This investment in AI-driven workflows comes during a tense period of belt-tightening. According to a recent Bloomberg report, Disney is preparing to lay off thousands of employees as part of a broader push to cut annual costs by 20 percent.
AI to the rescue of doing more with less? My guess is that Disney is chasing the holy grail: tapping into AI capabilities without having to invest heavily in developing them itself. And perhaps betting that Adobe will prove a more durable partner than OpenAI.
Graphx. Branded QR Codes for Event Photo and Video Sharing.
Graphx has launched Picsha Events, a tool that lets event guests scan a vendor-branded QR code to upload and share photos in an event gallery — no app or account required. The gallery can then offer a photobook of the event, along with prints and gifts from the vendor’s storefront.
The release joins a broader set of new and existing offerings under Graphx’s AI-powered Picsha platform, including automated photobooks, cards, posters, gifts, retouching, and white-label printing.
These services come together in the new Picsha Hub, which gives retailers a single branded webpage, order dashboard, and in-store kiosk home screen. Stores can fulfill orders in-house or through an outsourcing partner. Graphx offers the platform on a revenue-sharing basis, with no subscriptions or minimums.
Stanford, Radical Numerics & Paul Melcher. The Power — and Limits — of Image Captions.
As our industry friend Paul Melcher points out, new and better AI image generators tend to follow the release cycle of increasingly powerful large language models, or LLMs — the technology behind tools like ChatGPT.
But, as he also notes, a recent study suggests that high-quality image generation doesn’t require a massive LLM pretrained on text alone. Researchers from Stanford University and Radical Numerics introduced Halcyon, a family of generative AI models trained on image-caption pairs. Their findings highlight three points:
-
Text-only pretraining didn’t help. In the researchers’ experiments, pretraining on massive amounts of text alone did not improve image-generation performance.
-
Higher quality with less computing. The researchers report that Halcyon outperformed much larger LLM-powered image-generation models while using up to 275 times less compute.
-
The benefits worked both ways. With the same training budget, Halcyon also learned to generate accurate image captions.
Yet better image generation doesn’t mean generated images can fully replace photographs. The distinction goes beyond accuracy and trust — topics we discussed at length at Visual 1st — to what an image can contain.
As Melcher puts it:
“A generated image holds what somebody named: what you asked for, what a language model added to your prompt before the generator saw it, or what sat in the captions it trained on. (…) A photograph was made before any description of it existed, which is why it routinely holds things that were never requested, never noticed at the time, and never written down, and why those things stay in the file whether or not a viewer ever climbs down to find them.”
Pangram. Detecting AI-Generated Images and Videos.
Concerned about deepfakes? Provenance metadata and watermarks can help establish an image’s origins, but most images we encounter online lack that information. That leaves plenty of room to wonder: is this real, AI-generated, or somewhere in between?
This year’s Visual 1st Best Technology winner, Resemble AI, has made deepfake detection its mission. Now another contender is joining the field: Pangram, best known for detecting AI-generated text.
The company has announced Pangram Image 1.0, which it says offers:
-
Detection of images from leading AI image and video generators, with up to 99.96% accuracy on Pangram’s internal 1,200-image benchmark.
-
Heatmaps that highlight areas identified as AI-generated.
-
Resilience against common evasion tactics, including compression, blur, noise, and low-resolution scaling.
Pangram also introduced Image Flash 1.0, a smaller, faster model designed for frame-by-frame analysis of AI-generated video and deepfakes.
Both models support detection through a Chrome extension and offer API access for images and video.
Blurb. An Instant Store for Direct-to-Audience Book Sales.
Blurb, the self-publishing platform powered by RPI Print, has introduced Blurb Instant Store, a streamlined way for authors and photographers to create a standalone product page and sell books directly to their followers.
With no setup, platform, or distribution fees, creators set their own retail price, pay the unit printing cost, and keep the remaining profit. Blurb handles on-demand printing, shipping, sales tax, and customer service, so sellers don’t need to hold inventory or pack orders.
The tool also includes Blurb’s proprietary AI Listing Optimizer, which helps creators refine book listings across retail channels and generate keywords for SEO, marketing, and advertising campaigns.
BRIA. Background Removal, Version 3.0.
Past Visual 1st Best Technology Award winner BRIA has announced its third-generation image background remover, Bria RMBG 3.0.
The new model promises cleaner edges around hair, fur, and fine product details while preserving transparency in glass, smoke, and shadows rather than cutting them away.
RMBG 3.0 is available through Bria Create and BRIA’s Remove Background API, or it can be self-hosted in your own cloud environment.
Sundrop. Giving Baby Photos a Life Beyond the Camera Roll.
Sundrop is a new iPhone app that helps parents organize and enjoy the growing collection of baby photos sitting unused in their camera rolls.
The app scans iPhone photo libraries, filters out duplicates and screenshots, and automatically generates “drops”: weekly Sunday recaps, growth time-lapses, side-by-side sibling comparisons, and themed creations such as magazine covers.
Designed with privacy in mind, Sundrop processes images locally on the user’s device using Apple’s Vision framework and Core ML, rather than uploading them to external AI servers.
Parents can link their accounts to sync shared photos through Apple’s iCloud, manually edit creations, and automatically organize similar shots into galleries.
Sundrop costs $39.99 per year after a one-week free trial. Family Sharing lets both parents use a single subscription.
Best,
Hans Hartman
|