Tech Landscape #351
Snap’s new Spectacles, YouTube’s Communities, Amazon’s AI, and a whole lot more.
Hello!
Plenty of updates this week as we’re heading into the second partner conference season of the year. Snap and YouTube both held events, and Meta Connect is next week which I’m really curious to see.
Let’s get on with it. Hope you’re well!
Platform Events
Snap Partner Summit.
The headline announcement was a new version of Spectacles powered by Snap OS, but there was also news of many new AI features for creators and developers, and a forthcoming simplified app redesign.
The new Spectacles feature auto-darkening transparent LCD lenses with a 46° field of view, and Snap OS provides hand gesture control and environment mapping for mixed reality. Read The Verge’s hands-on for more.
The Spectacles are bulky and not very attractive, but they’re not aimed at the public; they will, in effect, be leased to developers. Snap doesn’t expect these to be a meaningful part of the business for another five years. Whenever new XR hardware is launched, the date they become mainstream seems a few years further away.
Made on YouTube 2024.
Pre-eminent among the feature announcements: Google DeepMind’s Veo video model will generate Dream Screen backgrounds and standalone videos; Creators can start Communities for their fans to gather, and the comments section will rebrand to Community Hub; Hype will let fans boost smaller creators; and Jewels are a micro-currency which can be used to send Gifts to live vertical video creators.
blog.google/products/youtube/made-on-youtube-2024/
Communities and Jewels are interesting to note: this is Discord territory.
Immersive & Spatial
Apple launched visionOS 2.
Key new features include: converting 2D images to spatial photos; more hand gestures; keyboard and mouse support; and support in Safari for viewing web-based XR experiences.
apple.com/newsroom/2024/09/visionos-2-for-apple-vision-pro-is-available-today/
I doubt this will be enough to convince any doubters to buy one, but it’s a nice enough update for existing owners, and a small step towards preparing the platform for the mass market. I wonder how much of this we’ll see make its way to Meta’s Horizon OS?
HTC launched the Vive Focus Vision XR headset, with colour video passthrough for XR and DisplayPort mode for streaming screens into the headset. prnewswire.com
It’s aimed at enterprise and high-end gamers, and the £1,000 price tag reflects that.
Synthetic
Amazon introduced AI tools for sellers.
They include a ‘personalized selling expert’ (chatbot) codenamed Project Amelia, generative video for ads, and improvements to product listings and recommendations.
aboutamazon.com/news/innovation-at-amazon/amazon-generative-ai-seller-growth-shopping-experience
Businesses are experimenting with putting AI in everything. Some of this might work, some of it might not.
Kling updated its video generator: the existing 1.0 model now has a motion brush for guiding movement, and the new 1.5 model produces higher quality 1080p output. x.com/Kling_ai
I made this example of Motion Brush in action:
Runway's Gen-3 Alpha Turbo model can now generate vertical videos. x.com/runwayml
Luma launched an API for Dream Machine and Runway launched an API for Gen-3 for developers to add video generation to their apps. The Gen-3 API is the more bare-bones of the two.
Also this week Runway announced a partnership with Lionsgate “centered around the creation and training of a new AI model, customized on Lionsgate’s proprietary catalog”. I highly doubt this will be used in movies themselves, but it seems an inevitable new step in animatics and pre-visualisation. It’s been reported that OpenAI is shopping Sora around content studios too. And while we’re all still waiting for Sora to get a public release, here’s an incredible example of what it’s capable of made by artist David Sheldrick.
Google is adding more image authenticity features, exposing C2PA metadata in synthetic images and adding metadata to generated ads. blog.google
Music generator Jen launched R3imagine, which can create a full track based on user-supplied audio. instagram.com
I haven’t managed to get a good result from it yet. The unfortunate truth is that ‘safe’ models, trained on more limited data, are simply not up to the standard of the ones trained more broadly on copyrighted material.Genspark released Autopilot Agent, an asynchronous AI agent that autonomously completes complex research tasks even after the user closes the webpage. mainfunc.ai
So you give it a statement, it breaks it down into component tasks, then goes off to multiple sources to check them out, repeating several times, then synthesises a result and emails you when it’s done. Take a look at this example I made. Really impressive and useful.
Social
Instagram introduced Teen Accounts which will limit the content they see and who can contact them, with parental controls. about.fb.com
All users aged under 16 will be automatically opted in, as will all new users aged under 18. This is Meta getting ahead of inevitable regulation.Threads posts can now contain up to 20 media attachments, matching the recent Instagram post change. threads.net/@threads
You know my theory that Meta wants to move the Feed out of Instagram so that it can become a fully video-first platform? This supports that.Substack added live video, allowing writers to host real-time video sessions with their subscribers. on.substack.com
Substack is slowly but surely becoming a social platform, rather than a newsletter platform.
Cool Things
Fortnite: Reimagine London, made for Zaha Hadid Architects to help young people have their say on the future of London’s public spaces and private buildings. I’ve seen this done in Minecraft before, but Fortnite adds a level of polish.
Everies combines Google Gemini with augmented reality to bring inanimate objects to life.