Chimera Vocal Synth
chimera-vocal-synth.vercel.appThere are a bunch of fun examples on Twitter of things people are doing with Chimera:
https://x.com/jamesonhodge/status/2118733265372860826
https://x.com/klausbrave/status/2119372853502177323
https://x.com/CupOJoseph/status/2119260104641900595
I came in cold and the first video made my day.
I made another:
That's awesome. Did you do that "by hand", or did you script Chimera to produce it?
I had Claude read a PDF of the arrangement and then spent a few hours tweaking the performance to match how it actually sounded.
Hilarious! Thanks for sharing :)
Yes, you did spend too much time on that, but it was worth it!
Great! Now barbershop quartets are unemployed! Will it ever end?!?!
:-(
Porn hasn't put sex workers out of business.
RIP White-collar-plus-bow-tie jobs
There are fart simulators too now. People are having fun with this. https://x.com/childlandlord/status/2107960660860502171
Shame there's no web-midi on this one though.
Yes but why do they all sound Indian? It's like hearing Apu singing :)
More like Stephen Hawking. Or the TTS shipped with a new Soundblaster cards in 1998. :)
Or the Voder from 1939
There does seem to be a bug - if you drag the tongue to the bottom left-most position, close the lips and bring them back as far as possible, raise the larynx all the way, and lower the soft palate all the way, the synthesizer starts making some kind of inhuman screeching sound (which is definitely not physically accurate).
My best guess for why is that the simulation parameters are set up slightly wrong in a way that is mostly fine for normal simulations (not an audible amount of distortion) but quickly goes off the rails on weird cases like the aforementioned.
Overall a really nice app though; honestly I was kinda blown away by it.
I could see this being very useful in a number of applications.
Being trans, of course voice affirmation training.
Then singing comes to mind, followed by language learning.
A solid visual demo like this would help give people more than "okay now more like the a in aunt"
The next step is clearly real-time MRI so you can more easily match the reference: https://seeingspeech.ac.uk/ipa-charts/
I pronounce it aunt but a lot of people pronounce it aunt.
Huh, that wouldn't have occured to me. Thanks for a new perspective!
I guess this is like a modern version of pink trombone - https://dood.al/pinktrombone/
Also is the name a Teto reference?
> https://dood.al/pinktrombone/
Discusses many years ago: https://news.ycombinator.com/item?id=18912628 and https://news.ycombinator.com/item?id=14135658
which is itself a version of research going back nearly 100 years.
e.g. the "Voder" by Homer Dudley - Bell Labs, 1939:
Was going to write about that too! The Wikipedia article is a fun read:
> This was a complex machine to operate. After months of practice, a trained operator could produce recognizable speech.
Yes, this reminds me exactly of pink trombone. Thanks for posting the link, I'll re-add it to my bookmarks.
This reminds me of that time researchers recreated the voice of the Pharaoh: https://www.youtube.com/watch?v=G8qhrURrQbI
There was a post to a tweet about this but I also found there's a GitHub repo for the software and this link to it running!
Earlier tweet: https://news.ycombinator.com/item?id=50410728
GitHub repo: https://github.com/jamesondh/chimera-vocal-synth
Scrolling is an issue on iOS. Where my finger lands to scroll the page is captured by the head view. Took me a while to figure out how to get passed it.
I learned to throat sing to put my baby to sleep. I'm having fun trying to figure out what all the shapes of the different mouth/throat components I make to get the same sound in the synth.
This is the kind of experience I want on a Monday morning
That was very intriguing, I was able to tune the false vocal folds and adopt the proper anatomical positions and it did output a proper Khoomei overtone. I wasn't able to find the Kargyraa range, but I managed a decent Sygyt.
Damned impressive setup.
I'm still trying to figure those out with my human vocal equipment. I'm impressed that you got this to do it!
I can't work out how Klaus did the Mr. Sandman version.
Is that a modded version with 4 quartet singers at once?
Maybe this one? https://github.com/klausbrave/yodel-studio
This is weird and fun. Brings me back to one undergraduate class where we used the IPA a lot
love the op-1 vibe ui
My work is done for the day. https://bsky.app/profile/stewleach.bsky.social/post/3mzts6gg...
Time to figure out that damn Australian "o" by flapping the tongue about...
Next step: reverse engineer Freddie Mercury's glottal physiology from queen tracks and fourier analysis, pass it around as a VST plugin.
Please and thank you.
I wonder if the Freddie Mercury License ("Do whatever you want with my music, just don't make it boring") meets the OSI Open Source Definition
Will make a great vocoder, once a mic is connected to it.
The vocal folds though? Lol
Is it not supposed to work on phone? All I hear is noise and engine keeps loading beyond 100%
It works on iOS for me
Can't whistle?
Not a vocalisation :)
The eye blink concerns me ...
Can it do throat singing?
Looks like there is no distortion in this simulation. Would be fun to play with different kinds of distortion as well
I love how creepy 3D view is.
Thanks, I hate it haha.