Thanks to Simon
Yew for alerting me to the AI powered, skybox
creation from Blockade
Labs .
https://lnkd.in/gTfHZDGw
I started with prompts along the lines of "large
urban space in 18th century London."
The first response was quite encouraging: a three
storey streetscape with a Georgian theme. It does feature a couple of modern
vehicles and it's a street, not an open square, so how will it blend with a
Revit project?
I adjusted the prompt to ask for an open space in the
centre. I got the open space but the camera is across to one edge. Worth
trying. Roll the dice and you get several modern buildings plus a grotesque
remake of Big Ben🙄 Hmm
I kept trying and... Hats off for inventiveness, but
generally speaking the architecture was better suited to an imaginary city than
18th century London.
Big Ben and Tower Bridge are definitely 19th century
and London is not really known for its continuous skyline of pointed Gothic
spires. Is Pugin's ghost trapped in the neural net perhaps?
One version has a raised camera height and wierd black
smoke belching out of many of the chimneys.
More compilations from my attempts
to generate 18th century London skyboxes for eg Project Soane using AI from
Blockade Labs.
A cross between tower bridge and Big Ben crossing a
street instead of a river. Lots and lots of spires. Buildings from Abu Dhabi 🤔 Often the same
"iconic" building features twice in the same image.
I tried pushing time back another century, then another. The ghost in the machine is very clever but it doesn't have a clear sense of history. Would work better for fantasy worlds probably.
Would any human being think that two Big Bens in the same image was acceptable? So it was kind of fun, but so far I don't think I have a skybox I could throw into Enscape3d for the Bank of England or similar. But I should probably give it a try.
Interesting insights into the strengths and weaknesses of this technology. Thanks Simon
Interesting to see posts from Benjamin Guler and Daniel Stine, AIA, IES, CSI, CDT, Well AP about bringing in texture mapped topo from Google Earth /Sketchup to Revit. The work flow for this certainly seems to have got easier since I attempted it over a decade ago. Not that I was the only person to work on this back in the day, but I do enjoy looking back at my early BIMpencil studies
So it was December 2011 and I was
attempting Gaudi's Parc Guell, of all things. Of course there was a lot of
cheating and gross simplification but as always my primary goal was to achieve
a deeper understanding rather than create a polished end product.
The process was quite demanding, especially scaling the
texture map to match the topography, but I sure had a lot of fun.
Along the way I learned a lot about both Revit and that
raised deck with wiggly seats around the edge that had made such a huge impact
on me in the late 1970s when I went on my Gaudi pilgrimage.
Post 3 about my 2011 study of Parc
Guell.
This is the closest I got to convincing images that
blend the Revit model of the "building" with the textured toposurface
context.
The piece of topo closest to the raised deck is a grass
material surface with subregions in a white material to represent paths. The
rectangular building is/was a school, and when I visited 45 years ago,
uniformed kids came out to play on the deck while we were there.
I'm still rather proud of the upper image of the
columns supporting the deck, and the stupendous view over Barcelona towards the
sea. Native Revit rendering of course. Happy memories of the little square
boxes jumping around the screen as the image came into sharper focus.
To say it was tedious by today's standards seems a
banal understatement.
No comments:
Post a Comment
I've been getting a lot of spam so had to tighten up comments permissions. Sorry for any inconvenience. I do like to hear from real people