Half a croissant, on a plate, with a sign in front of it saying '50c'
h a l f b a k e r y
Huh?

idea: add, search, annotate, link, view, overview, recent, by name, random

meta: news, help, about, links, report a problem

account: browse anonymously, or get an account and write.

user:
pass:
register,


                           

Speak & 3D-Scene-Constuct

'Computer, three men and a boat, please'
  (+2, -1)
(+2, -1)
  [vote for,
against]

Bit of a long-shot, considering how long it's taken for get working speech-to-text, but imagine being able to tell your workstation what you want your scene to look like, then direct the props, lighting and affects, all by voice.

Obviously, you'll need hands-on to make things for your scene, but the rest could be sped up considerably if the computer knew what 'make little Johhny stand behind the tree', 'have Johnny throw the ball to Jenny', or 'set Johnny on fire' meant....

MonTemplar, Aug 19 2001

Winograd's SHRDLU http://www.cs.cf.ac...I1/COPY/shrdlu.html
"Put the red block on the green cube." Pretty advanced for 1972. [rmutt, Aug 19 2001, last modified Oct 04 2004]

Microsoft Research: ChatPainter (2018) https://arxiv.org/pdf/1802.08216.pdf
Sharma, Suhubdy, et al: ChatPainter: Improving Text to Image Generation using Dialogue [jutta, Sep 02 2019]

Put That There (1979 - basic demo) https://www.youtube...watch?v=RyBEUyEtxQo
Another system for moving shapes around by voice commands and optionally pointing [notexactly, Sep 07 2019]

Put That There (1980 - advanced demo) https://www.youtube...watch?v=sC5Zg0fU2e8
More interesting scenario, interactive clarification, and two users at once, with the ability to protect shapes from other users' commands. Also supports object naming and painting by pointing as seen in video PeWwfhsSqsc, as well as destruction animations and alternative voices as seen in -bFBr11Vq2s. Comments disabled :( [notexactly, Sep 07 2019]


Please log in.
If you're not logged in, you can see what this page looks like, but you will not be able to add anything.



Annotation:







       Yes, we'd all like computer who understands what we want and just does it. (The halfbakery term for this kind of thing is WIBNI, from Wouldn't It be Nice If.)   

       The hard part of this is understanding what it means for little Johnny to stand behind the tree, and how little Johnny's posture and facial expression changes as he peeks out from behind the tree.   

       Having a computer parse the word "behind" and move a cardboard cut-out of the object "Little Johnny" behind it is trivial and has been solved for about 30 years, as [rmutt]'s link points out.   

       (A small version of "blockworld" was one of the programs we wrote in my first Lisp class 20 years ago. Ah, memories.)
jutta, Aug 27 2001
  

       Students of Prof. Sheldon Klein "bake" this every year in CS774 (Multimedia and Virtual Reality). The problems involved are truly interesting--we did this in a Virtual Reality environment and spent most of our time teaching the computer concepts of "behind" "in front of" etc, also had a pretty powerful rule system (i.e., if tree is in the way, ball can't be thrown, etc.)   

       Croissant for independently coming up with the idea, though...   

       http://www.cs.wisc.edu/~sklein/sklein.html
cowtamer, Apr 18 2003
  

       ive been toying with such an idea. i was thinking of reversing image recognition algorithms somehow.
ironfroggy, Apr 18 2003
  

       Le'see here, what would this be like . . . .   

       "Create primitive, box, and name 'cube one'. Place in current view. Move cube one to X, zero, Y, zero, Z, zero, using current view coordinate system. Re-center cube one local axes to centroid of cube one, affecting pivots only. Move cube one to X, zero, Y, zero, Z, zero, using current view coordinate system. Add texture mapping coordinates to cube one. Convert cube one to polygon mesh. Tessellate cube one sub object polygon one. Tessellate cube one sub object polygon one. Tessellate cube one sub object polygon one. Soft select, range twenty, pinch zero point 5, bubble zero point seven five, cube one, sub-object pre-tesellation polygon one lower left corner vertice, ignoring backfacing vertices. Name vertice lower left corner. Move lower left corner, relative to adjacent polygon normal  . . . ."   

       Hmm. Guess it'd only work with pre-built models or a digital backlot.
bristolz, Apr 18 2003
  

       I've added a link to a relatively recent Microsoft Research paper that gives you some idea where the industry is in terms of understanding how people talk about scenes.
jutta, Sep 02 2019
  

       // 'set Johnny on fire' //   

       [+]   

       "Buld a man a fire, and he'll be warm for a day; set a man on fire, and he'll be warm for the rest of his life ... " (Pterry Pratchett)
8th of 7, Sep 07 2019
  

       When I was little, I had among my toys a little dump truck. It was not very notable, apart from what it said on the front above the windshield: "CONSTUCTION TRUCK".
notexactly, Sep 07 2019
  

       Presumably that means that the buildings aren't welded or bolted, but stuck together ...
8th of 7, Sep 07 2019
  

       CONSTUCT: what happens when the inmate tries to leg it out through the drainpipe at Alcatraz
Sgt Teacup, Sep 07 2019
  


 

back: main index

business  computer  culture  fashion  food  halfbakery  home  other  product  public  science  sport  vehicle