"Bare-handed speech synthesis"
« previous post | next post »
This is neat: "Pink Trombone", by Neil Thapen.
By the same author — doodal:
March 22, 2017 @ 6:38 pm · Filed by Mark Liberman under Computational linguistics
« previous post | next post »
This is neat: "Pink Trombone", by Neil Thapen.
By the same author — doodal:
March 22, 2017 @ 6:38 pm · Filed by Mark Liberman under Computational linguistics
Powered By WordPress
AntC said,
March 22, 2017 @ 11:05 pm
Thanks Mark. Neat and so much fun! (on an iPad)
The 'User Manual' is here https://m.youtube.com/watch?ebc=ANyPxKpTGoty36n22nakl55Y971bPSLbpUkPNVPiq1cJ5_pSLDBL7W9YcYrZm5xOPi17rx_w75lwpg_65wWYeZVmvaM58_jatA&v=7LGnozlwU1o
Even after watching that, I can produce speech-like sounds, but not anything I could recognise as connected language. It just needs too much controlled coordination.
How babies could ever learn to control all those bits is beyond me.
[(myl) Babbling! ]
Adrian Morgan said,
March 23, 2017 @ 8:12 am
Re Doodal: I wonder how many people remember Dick Oliver's "Fractal Vision" (or "Fractal Grafics" as other versions were called) back in the early 90s. I still have my copy (which runs in a DOS emulator) as well as the 485-page manual.
Dan Lufkin said,
March 23, 2017 @ 9:34 am
The fractal nature of language would be an interesting topic to kick around for a while. There are a few Ghits along that line, but the idea doesn't seem to have affected main-line linguistics much. (Now watch me get savaged.) Zipf's Law suggests that there's something deep going on but so does Bode's Law (qG). I recall that Noam Chomsky could demonstrate that Zipf's Law was an emergent property of logarithms but I can't reconstruct the details.
I'm glad to see Fractal Vision mentioned and fondly recall Fractint and the Stone Soup Group and Rollo Silver's Amygdala newsletter. Fractals inspired a lot of clever programming a generation ago; perhaps Doodal will spark some new concepts. It looks promising.
Matt McIrvin said,
March 23, 2017 @ 9:35 am
It seems to me that Pink Trombone needs teeth. It can't say the consonant D very well.
[(myl) It would also be good for it to be able to store and replay time sequences, in a format that could easily be manipulated as the output of an articulatory synthesis program.]