Tech Landscape #336
AI Music takes another leap forward, Meta Quest gets a quality update, and Humane's AI Pin lands with a thud.
Hello!
An important update to start with: I’m going on holiday, so the newsletter will be taking a break for a few weeks. The next issue will most likely arrive on 13th May, at which point there will have been four weeks’ worth of news to catch up on, so spare me a thought as I plough through it all and try to make sense of it.
Alright, let’s get on with it. Hope you’re well!
AI & Synthetic Media
Udio is a (stunning) new song generation tool.
It can generate up to 30 seconds of a song, with vocals, that can then be extended forwards or backwards, with high quality results.
x.com/udiomusic/status/1778045322654003448
The output is astonishing. As in, very difficult to tell apart from ‘real’ songs. I suspect it’s been trained on scraped copyrighted music, because it’s so good and the diversity of styles is so broad; and I suspect that’s going to get the music industry’s hackles up. So enjoy it while you can. Here’s a short clip I made about falling in love with AI, in an old school hip-hop style.
Ideogram upgraded its 1.0 model to improve text rendering and photorealism, along with new features including a Describe function and negative prompting. about.ideogram.ai
Ideogram’s model is really good; the results are consistently strong, and it has the best text rendering. But it’s not as effortlessly stylish as Midjourney, and not as flexible as Stable Diffusion, and not as “safe” as Adobe Firefly, and so I wonder if it can survive long-term as a standalone product.
Image editing app Facet added Compose, which enables multiple prompts in different regions of a single image. threads.net/@facet.app
You can prompt different regions which can be moved, resized, and edited, then create variations which preserve the editable regions (take a look at my video below to see what I mean). This is another interesting workflow experiment from an independent app.
Open AI released an upgraded GPT-4 Turbo model with computer vision capabilities, available through the API¹ and rolling out to ChatGPT². x.com/OpenAI¹, x.com/OpenAI²
ChatGPT will gain a Dynamic version switching option, automatically changing models depending on the user’s requests. threads.net/@luokai
Meta is beginning to test Meta AI outside of the US, in India and some African regions, through Instagram, Messenger, and WhatsApp. techcrunch.com
Still not in the UK, though (like a lot of Google AI products too). I wonder if this is just home market preferencing or if there’s some element of the regulatory regime here in the UK that makes them harder to launch. It can’t be a language.
Google Cloud Next
Google held its annual Cloud Next event, with a lot of announcements for enterprise customers and a few that are relevant to our interests.
Gemini 1.5 Pro is available to developers with new features including speech understanding and custom instructions. developers.googleblog.com
It has a huge context window and can analyse audio now so you can, for example, upload a podcast episode and ask a bunch of questions about it. Quite incredible.Imagen 2 can now generate 4s motion clips, and also gained image editing capabilities. cloud.google.com
Vids is a new video creator coming to Workspace which uses AI to create a storyboard from Slides, generate scripts, choose stock video, and more. workspace.google.com
After years of owning the world’s biggest video platform in YouTube, Google is finally getting serious about building video creation tools too.
Humane AI Pin
When the AI Pin was announced last year I said I’d hold back my opinion until I’d read the reviews. Well, the reviews are in, and… oof. Here are a couple of the kinder ones:
Should you buy this thing? That’s easy. Nope. Nuh-uh. No way. The AI Pin is an interesting idea that is so thoroughly unfinished and so totally broken in so many unacceptable ways that I can’t think of anyone to whom I’d recommend spending the $699 for the device and the $24 monthly subscription.
I think the AI Pin in particular is built on the premise that people don’t like their phones, and I don’t believe that to be true. Sure, you might feel like you use some social apps too much; but phones are just unbelievably useful for so many more things.
I wondered if there was much sense in making a $700 phone replacement, but it makes no sense at all if it can’t even get the basics right. I still think there’s a potential new device type in here somewhere; let’s see if the Rabbit R1 can do better.
Social & Messaging
YouTube expanded its shopping features with curated Collections, bulk tagging, and the Affiliate Hub. blog.youtube
TikTok has been captured by shopping, and largely by cheap drop-shipped imports. I hope YouTube can do it better.Messenger allows sending HD photos, along with creating shared albums for group chats, and sending large files. about.fb.com
Instagram fully rolled out Cutout Stickers, enabling regions of images and short video clips to be used in Reels and Stories. threads.net/@instagram
Meta announced that the Threads API is coming soon. It’s currently in testing with select partners, and developer documentation is now available ahead of a general release in June. developers.facebook.com
I’d like to see more start-ups and businesses on Threads; perhaps reducing the friction of cross-platform posting will make that happen.Automattic (owners of Wordpress) bought Beeper, a cross-platform messaging app. blog.beeper.com
Automattic also recently bought texts.com, which does a similar job in a desktop app, and the teams will be brought together on a unified product. But do people want another messaging app? Especially one that doesn’t have all the individual features of each? We’ll find out.
XR & Spatial
The latest Meta Quest update improves passthrough quality on Quest 3, along with support for external microphones. meta.com
The video passthrough on Quest 3 is already pretty good, this apparently improves the detail, making it easier to, for example, read your phone. I’m eagerly awaiting this software update to roll out so I can try it for myself.Meta added Instant Replays to Horizon Worlds, to make it easier to share gameplay moments. meta.com
Improve the popularity of the platform by improving discovery by improving the ease of sharing highlights.Luma Labs launched an Android version, of its 3D scanning app. x.com/LumaLabsAI
Luma makes impressive 3D scans; this is an amazing feature, but is it enough to be an amazing product?
Everything Else
Roblox is getting in-platform programmatic video ads, in partnership with PubMatic. newdigitalage.co
eBay added "Shop the Look," which uses AI to suggest similar and complementary outfits based on users' shopping history. innovation.ebayinc.com
Brands & Entertainment
Kung Fu Panda: School of Chi is a guided meditation spatial app for Apple Vision Pro • Niantic’s Peridot is appearing in SunnyTune, a fun weather app for Apple Vision Pro. It’s nice to see a few apps breaking out of Apple Vision Pro’s windowed interface.
Coca-Cola is teaming up with Marvel on customised cans with digital enhancement. Fun trailer.
The Faceless Lady is a serialised live-action horror story in Meta Horizon Worlds, presented by moviemaker Eli Roth.
Roblox publisher GeekOut K.K. is offering a $1.5 million fund for creators to make Attack On Titan content, with the winners also being granted an official license • The movie Godzilla × Kong: The New Empire launched a ‘playable trailer’ in Roblox, with unlockable limited edition merch • Nike returned to Fortnite with Airphoria 2, a redesigned limited-time experience plus items for sale in the shop • Coachella is coming back to Fortnite with a backdrop and new tracks in Festival, plus outfits and emotes in the shop.
Weirdest team-up? Modern Warfare 3 and ’70s stoners Cheech and Chong.
NFT character franchise Doodles announced Dullsville and the Doodleverse, a “storytelling experience” featuring music by Pharell Williams and the voice of Lil Wayne.