While I’m enjoying these Blender posts everyone is doing, I’m confused on why it’s blowing up NOW and not back when Seedance 2.0 came out, it was widely known back then, just most people don’t know how to use blender (I do)
I swear this entire world is a fking joke.
OpenAI named its 3 new models Sol, Terra, and Luna.
So… somewhere inside OpenAI, a crypto intern put the 2022 collapse starter pack into a naming doc as a joke.
Then 14 directors, 6 branding people, 3 lawyers, and one guy named Chad from product all said:
BEAUTIFUL. Ship it. Yea, this feels like the future.
Is someone trying to warn us about the post-IPO price action?
As I’ve been saying!
Ok nowhere near this eloquently.
I think this has interesting consumer repercussions: when anyone can generate anything, these realtime 3D scaffolds empower them with control in a fun, game-like experience rather than complex timelines and editors
for each level of media richness, there is a deterministic scaffold we lean on until the generative layer is fully capable on its own. for text, it was the python interpreter that stopped llms from hallucinating math. now that video conditioning is a thing, it is 3d rendering engines that not only enable control but also address the long-standing spatiotemporal consistency problem in world models
the question is when this scaffold should go away, if ever (python is still here even as llms got smarter because we've found it to be useful beyond math verification)
Incredible to see what a professional animator can do with AI video.
He created a 3D previz and then had Seedance render the actual anime - with the motion and camera control preserved!
(made by @craftcapitallab)
This is what the next generation of experiences looks like.
One sentence becomes a quest, a world, and a cast of characters with real dispositions and rules, all generated in real time on Reactor's infrastructure.
Alakazam built the orchestration layer that turns a world model into something you can actually win or lose.
Full technical breakdown in the article below.
I hate AI video, however, this is directionally the way AI video needs to go.
Trying to endlessly generate prompts and roll the dice is awful.
Generating 3D scenes and allowing the user to pick the camera angles, lenses, depth of field, movements, etc, is how AI video needs to be done.
How long until I can dictate the exact descriptive text above to Siri, and have it understand and piece together those steps into a functional shortcut?
Completely agree.
The way you turn AI into a compounding asset is by codifying your best ways of working and then democratizing it across your company in a way that wasn't possible before.
One of the more common requests we get from companies earlier in their AI journey now is to stand up their internal skill library & fill that library with skills that we build through interviews/sessions with top performers.
i’m working on this too. iphones tend to crash at around 200MB VRAM which is such a banger of a limit that i rearchitected my entire world pipeline to aggressively stay under this limit.
most games have toggles/quality settings and if your fps sucks you lower them, but since ios safari just straight up crashes we can’t do that.
People aren’t ready for the fact that within five years, they’ll be able to drop themselves into their favorite movie or show, talk to beloved characters in real time, alter the narrative, and create their own stories with fidelity indistinguishable from the original production