Tech Landscape #416
Kling’s milestone release, Alexa goes States-wide, and handbags from Anthropic and OpenAI
Hello!
This was a quiet week; I don’t know why tech companies seem to hold things back in the run up to the Super Bowl, but it was the same last year.
So… I don’t have much to say. The (paid) subscriber-exclusive content is still in the works, if you’re interested; I’ve written a first draft, now I have to make it readable.
Anyway, let’s get on with it. Hope you’re well!
Synthetic Audio-Visual
Kling 3.0
Just a few weeks after its last big launch week [TL 411] Kling is back with the 3.0 series.
VIDEO 3.0 provides multi-shot generation at up to 4k quality with native audio, from a variety of image, video, of text references, with improvements to image realism (see below) and prompt adherence. It unifies its previous video models (2.6 and O1). app.klingai.com
It’s a big model that requires a lot of compute, so generating videos is currently slow. I did manage to run a few tests ⬇️; you can see that its prompt comprehension, the consistency between shots, and the visual quality are excellent, but it struggled with motion and physics. I’ve seen some amazing clips made with VIDEO 3.0, but haven’t had much success yet myself.
IMAGE 3.0 enhances cinematic realism through improved lighting and texture reconstruction, in up to 4K resolution, from multiple reference images, and has a new Image Series Mode to quickly generate variations for storyboards and more. app.klingai.com
It’s only available to paid subscribers so I haven’t had the chance to test it yet.
Creator Tools
The Kling Web app added Kling Agent, a creative assistant for task automation such as storyboard and video generation, smart editing, and batch creation. app.klingai.com
Freepik’s List Nodes in Spaces enable batch generation of shots, formats, text copy, and more. x.com/freepik
Adobe Firefly is offering unlimited image generation, including models from its commercial partners, until 16 March, and unlimited use of the Ray 3.14 video model for two weeks.
This isn’t an ad, it’s just a good deal.
Riverflow 2.0 is a new image generation and editing model aimed at creative and marketing teams, offering “reliability, font control, and detail preservation”. riverflow.ai
It comes from Sourceful, a brand design company / platform based in Manchester (England).
Avatars & Voice
Hedra introduced Omnia, a unified AI model for generating avatars / talking characters with a high degree of expression. hedra.com/blog/
ElevenLabs’ flagship V3 voice model is generally available, promising greater stability and accuracy, especially in reading numbers and symbols. elevenlabs.io/blog/
Assistants
Amazon’s Alexa+ is available to all users in the U.S. priced at $19.99 per month (or free to Prime subscribers) for use on all devices, and with a new free tier available in the Web and mobile apps. aboutamazon.com
Coding
This is largely outside of my area of interest, but it was a quiet week so I’ll take what I’m given.
OpenAI released the Codex app, bringing its agentic coding interface to MacOS, and introduced GPT-5.3-Codex, a new model that’s highly optimised for agentic coding. Anthropic released Claude Opus 4.6, an updated foundation model with optimisations for agentic coding (and finance).
These announcements were made within hours of each other, and in an indication of the growing rivalry between the two companies Anthropic also launched a series of video ads (e.g. this one) clearly targeting ChatGPT for its forthcoming inclusion of ads. Leaving aside the casual rip-off of Black Mirror and the misrepresentation of how ads will show up in ChatGPT, I don’t actually think Anthropic’s ads were very useful to a broader audience who don’t follow this stuff closely, and they underestimate how many people will accept ads as a value exchange. I also don’t think that Sam Altman laughed.
Social & Creators
YouTube’s auto dubbing is now globally available with an expanded library of 27 languages, and Expressive Speech which can “capture a creator’s original emotion and energy” in eight languages. blog.youtube
Substack is adding more theme customisation, moving away from homogeneous design to allow creators’ themes and branding to be more prominent, starting in the iOS app. on.substack.com
Everything Else
Roblox launched a beta of 4D generation, enabling generation of functional 3D models with interactivity, using its Cube foundation model. about.roblox.com
It’s starting with cars; so you can describe the vehicle you want with a text prompt and it will generate a fully-functioning car that’s ready to drop into games.