Tech Landscape #431
Meta wants you to pay up, Microsoft wants model autonomy, and Google wants to keep links alive.
Hello!
This was a very quiet week for news, presumably because there was a public holiday in the USA. I’m currently recuperating from a minor surgery and I don’t have anything useful to say, so I’m just going to get on with it.
Hope you’re well! (I am.)
Synthetic Audio-Video
Microsoft announced MAI-Image-2.5, hailing it as “a step change in quality over our previous model” with “major improvements in text rendering, stylised illustration and commercial imagery” and better prompt coherence and reasoning.
It’s set to be released soon although is already available on Arena if you have the patience to roll the dice a few times to see it (I haven’t managed to yet). What’s interesting here is the pace at which Microsoft is upgrading this model; version 2.0 was only released a couple of months ago [TL 422]. Clearly the company sees having its own image model as a priority.
ElevenLabs’ Dubbing v2 model preserves the original speaker’s emotion, tone, and pacing for more natural dubbing in over 90 languages. elevenlabs.io
ElevenLabs’ Music v2 model brings improved vocals, instrumentation, and arrangement, multilingual support, and section-by-section composition and inpainting features. elevenlabs.io
I haven’t tried this yet (I don’t have a paid account) but it’s trained on licensed data so is a commercially safe option for brands and businesses.AI music startup Mureka launched Co, a desktop app that integrates directly with tools such as Ableton. instagram.com/mureka.ai
Creative Tools
CapCut launched Design Studio 2.0, a redesign of its creative workspace with a combined agent and canvas interface. instagram.com/capcutapp
Tencent announced Miora, a creative design studio with a combined agent and canvas interface, for an international audience. tencent.com
Many (most?) creative tools are converging on this agent + canvas interface; the problem they face is: where’s the differentiation?
Higgsfield Reframe can automatically outpaint videos to fit new formats, with image references for improved consistency. threads.com/@higgsfield.ai
OpenArt added Characters with Voices for vocal consistency across generations. instagram.com/openart_ai
Runway MCP lets users generate images and videos directly within compatible AI agents like ChatGPT and Claude. runwayml.com
Social
Meta is launching subscriptions to Facebook, Instagram, and WhatsApp. Priced from $49.99 per month (Plus) to $499.99 per month (Max), they include account verification and features such as improved search ranking, links in Reels, and exclusive stickers. Future updates will include plans for Meta AI, and bringing them all together in an all-inclusive Meta One subscription.
The way this was announced was a little weird; just a video from Meta’s Head of Product, Naomi Gleit, with no supporting press release. Further details of the plans were released by users as they encountered them.
YouTube is testing custom feeds that can be generated from a text prompt. support.google.com
Currently rolling out to users in the USA.YouTube will show AI disclosure labels in more prominent positions on videos and Shorts, with automatic detection to label undisclosed photorealistic AI content. blog.youtube
Threads users can add music clips to a post or reply, up to 30 seconds in length with optional lyrics. threads.com/@threads
If you’re looking for more opinionated insight into the world of social you might want to check out VCCP Social Clubstack, a new semi-regular email newsletter brought to you by my friends and colleagues at VCCP’s Social Club.
Assistants & Search
Google improved quality content discovery in its AI Search features, adding Preferred Sources to AI Overviews, new perspective carousels, and a “Highly Cited” badge to highlight primary reporting. blog.google
Anthropic launched Claude Opus 4.8, an upgraded AI model with improved reasoning and honesty and “effort control” that lets users adjust response depth and speed. anthropic.com
Sesame launched a preview of its personal agents, a voice-first iOS app featuring low-latency lifelike characters with individual memories, real-time search capabilities, and an incognito mode. sesame.com
Microsoft 365 Copilot was redesigned with an updated interface and faster user experience, an expanded workspace prompt line, and single entry point across productivity apps. microsoft.com/microsoft-365/blog
Tencent Cloud launched WorkBuddy, an agentic workspace for complex office work and productivity. tencent.com
OpenAI’s Codex added computer use on Windows, letting it operate any installed app. x.com/OpenAI
Perplexity Computer is now available in Microsoft365. threads.com/@perplexity