Co-founder & CEO, Craftology: the context layer for media & entertainment. AI agents run animation production. 10 yrs of VFX & virtual production pipelines.
For years the advice to a lot of game devs was: show early, show often. We would use first playable / demos as our marketing and our market tests.
Now a demo is also a blueprint. A viral wall-punching game had a Roblox copy live in 25 hours, before its Steam page even existed. Alliance Arts canceled the public demo of Drunken Goddess Reflux days before Next Fest, worried AI could replicate it before launch.
So the question for us as an industry: what do demos, early access and pre-release marketing look like when showing means handing over?
And who builds the tools that find the copies, and prove who made it first?
#GameDev #AIinGames
Love this one from Embark. In THE FINALS, the AI commentator now calls you out by name, live, when you wipe a team or clutch a cash out.
The lines come from the game's context in real time, mixed with pre-recorded ones. They're starting with a few events so it still sounds like Scotty. Streamer and anonymous modes are respected.
A lot of studios stay quiet about using AI. Embark puts it on stage. Intriguing what their TTP will be 🙂
https://t.co/UTXyexY4Pn
#TheFinals #GameDev #AIvoice #GenerativeAI #Gaming
Suggest to check out Hy-Nomas, from Tencent Hunyuan. Open source, Apache 2.0.
An interactive video world you direct turn by turn. Type what happens next or move the camera, and it keeps generating the same world, segment by segment, with memory carried between turns.
Runs on a single RTX 4090 (INT8 weights, CPU offload). Training code is included, so you can fine-tune a LoRA on your own footage.
No benchmarks yet, the tech report is coming. But memory across turns is the part that matters for longer scenes.
https://t.co/jSAt2XHlDO
#WorldModels #AIvideo #GenerativeAI #OpenSource #AIanimation