A.I. Alert - thispe...
 
Notifications
Clear all

A.I. Alert - thispersondoesnotexist  

  RSS

The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
12/04/2020 7:11 pm  

Not related to AC but related to some of the discussions on the "Borg" and the scary possibilities of artificial intelligence taking over. Might be old news to some but its only now come to my attention, a year late.

None of these people are real:

faces

 

https://thispersondoesnotexist.com/

(refresh the page to see a new face)

A.I. has now got to the point where it can produce an image of a face that's indistinguishable from a real one. It throws the occasional glitch but for the most part these faces are pretty impressive.

More info on this:

https://www.theverge.com/tldr/2019/2/15/18226005/ai-generated-fake-people-portraits-thispersondoesnotexist-stylegan

 

Its like one of the "planes" or "world of forms" now intersects the mundane plane in a way it didn't before.


Quote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
12/04/2020 8:21 pm  
Posted by: @duck

A.I. has now got to the point where it can produce an image of a face that's indistinguishable from a real one

I wrote about this kind of stuff in Silicon Exposé a few years ago. From time to time, I drop a reminder about the "inorganic beings," the silicon-based consciousness that is alien to, and aggressively invading, us carbon-based units.

Although the insidious, deeper aspects of AI are not "artificial," they are presented to us mortals as extensions of our wit and smartness. The deeper layers ...

"Tetrahedral coordination is a major structural motif in silicon chemistry just as it is for carbon chemistry. However, the 3p subshell is rather more diffuse than the 2p subshell and does not hybridise so well with the 3s subshell. As a result, the chemistry of silicon and its heavier congeners shows significant differences from that of carbon ..." -Wiki

Carbon (us) and Silicon (chips, AI, tetra-archons) don't get along, but here we are, depending on it.

Posted by: @duck

Its like one of the "planes" or "world of forms" now intersects the mundane plane in a way it didn't before.

Exactly.

 


ReplyQuote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
12/04/2020 8:40 pm  

Oh, I forotted to mention that certain posters right here on LAShTAL have been identified as 'bots or humans that have been implanted with fixed, crystallized thoughtforms and agendas. We do not pray for them. Pity not the fallen.

 


ReplyQuote
 Anonymous
Joined: 51 years ago
Posts: 0
12/04/2020 8:49 pm  

https://www.cnbc.com/2020/03/03/bluedot-used-artificial-intelligence-to-predict-coronavirus-spread.html

A little after midnight on Dec. 30, artificial intelligence platform BlueDot picked up on a cluster of “unusual pneumonia” cases happening around a market in Wuhan, China, and flagged it.

BlueDot had spotted what would come to be known as COVID-19, nine days before the World Health Organization released its statement alerting people to the emergence of a novel coronavirus.

AI isn't necessarily an enemy, as histrionic conspiracy theorists believe.
 
Posted by: @duck

Its like one of the "planes" or "world of forms" now intersects the mundane plane in a way it didn't before.

There was some discussion about  AI being able to solve II:76 proving itself 'strong' and the child from no expected house.


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
12/04/2020 9:41 pm  

I was also impressed about the Jordan Peterson voice generator that they shut down after a while it went live.

https://www.notjordanpeterson.com/

 

It still had the hollowness to the sound, but clearly sounded like Peterson (I'm not a fan!).

 

But yes, the real concern is that when you can do convincing deep fakes of alive and dead people (already seen pretty good ones) and replicate the sound to a great degree you will be able to do extremely questionable things with that stuff.

 

I mean in 10 to 15 years there probably is some pretty advanced deep fake programs available you can feed a neural network with some of Crowley's recorded Aethyr calls and get the DeepFake analyze all the available photos of Perdurabo and you will be watching a "live feed" with a very convincing vision & the voice, perhaps even recite the Book of the Law in real time ( The second beast was given power to give breath to the image of the first beast, so that the image could speak  [Revelation 13:15] )?

 

Then again, there is certain hollowness to the sound with computer software especially when it comes to generating or mimicking the human voice and analog sound in general. I've been working with studio musicians, recorders and there is still some magic to real analog saturation that simply cannot be replicated as of yet. Most guitarists agree with me when comparing even the best Kemper profiling digital amps and other expensive digital modulation amps (that try to simulate the real thing of analog saturation) simply do not come close to the real thing (especially if you compare distorted electric guitars recorded through a real tube amp versus the best Kemper digital profiling amps). I still hear the difference. In 20 years I probably don't.

 

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
13/04/2020 3:14 am  
Posted by: @shiva

We do not pray for them. Pity not the fallen.

Well, "every man and every woman..." (and the Law is for all). No need to pity the fallen if one had 38 hands to to pick them up again. 😉 😉 😉 

 

Posted by: @xon

There was some discussion about  AI being able to solve II:76 proving itself 'strong' and the child from no expected house.

That's a nice one and almost comical in a way. If you happen to find that thread let me know. I plan on having a go with one of these AI algorithms as an experiment with AL. I will reveal it in time.

 

Posted by: @behemoth

feed a neural network with some of Crowley's recorded Aethyr calls and get the DeepFake analyze all the available photos of Perdurabo and you will be watching a "live feed" with a very convincing vision & the voice

This ties in with this thread from a couple of months ago:

https://www.lashtal.com/forums/ancient-egypt/mummy-returns-voice-of-3000-year-old-egyptian-priest-brought-to-life/

Here they are building an actual physical reproduction of a Pharaoh's voice box. Its also an example of these "planes intersecting".

 

Posted by: @behemoth

The second beast was given power to give breath to the image of the first beast, so that the image could speak  [Revelation 13:15]

This is brilliant! 😀 Top marks, sir. There was a recent thread on the "second beast" in case you missed it:

https://www.lashtal.com/forums/thelema/the-second-beast/

We never got to a convincing explanation for who or what the second beast could be, your explanation here is the best one I've heard!


ReplyQuote
 Anonymous
Joined: 51 years ago
Posts: 0
13/04/2020 4:39 am  
Posted by: @duck

No need to pity the fallen

It's wise to be careful casting aspersions about others being fallen. Always question why such things should or would be said about another.  It's often better to look close at oneself and question whether the accusation is a means of avoiding confronting and destroying one's own illusions about oneself.

A magical partner once quoted this in regards to me when some were saying how fallen I was, it does apply to all Thelemites.

they that see thee shall fear thou are fallen

We laughed at their fear.

Posted by: @duck

If you happen to find that thread let me know. I plan on having a go with one of these AI algorithms as an experiment with AL

Link . It was mostly my strange musings from offline conversations.

I put II:76 into some online AIs around that time and got back some funny responses. One was kinda interesting and the response is still in my files:

Q: 4 6 3 8 A B K 2 4 A L G M O R 3 Y X 24 89 R P S T O V A L What meaneth this, o prophet?

A: It means, it is God that gives us all these commandments.
(5) And now we know the law is of God, and has nothing of it but his commandments: but this is all it is.
And now we know that every thing which is in heaven was given by him to be given us, who being in heaven, see no need of them; and they are given to his Son, which is without beginning, no power, not even the power of God.
And now the law hath no force except from the one, whose obedience, and keeping of it, is of the Lord Jesus Christ.
(6) Now no one can come to the Father except the Son; who is the firstborn of every creature, and of all things in heaven; and all things on earth were made through him, and for him; which he hath given us, that we should walk in them.
And now the Son of man hath not come from God without a commission from him; the devil hath.

Looking forward to reading what you discover with the A.I. experiment.


ReplyQuote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
13/04/2020 6:04 am  
Posted by: @duck

I plan on having a go with one of these AI algorithms

Duck vs Goliath. You see how history is constantly repeating itself?

 


ReplyQuote
djedi
(@djedi)
𒇽𒅗𒅗𒈠
Joined: 9 months ago
Posts: 335
13/04/2020 6:34 am  

Here's one it gave me:

thisperson

If you run through enough of these fake faces, a common quality starts to bleed through. An astonishing number have this FAS look about them, but that's not it.


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
13/04/2020 6:55 am  
Posted by: @djedi

An astonishing number have this FAS look about them, but that's not it.

 

I believe it's the NPC syndrome. Some of them also have the so-called Fluoride Stare (controversial).

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
christibrany
(@christibrany)
Yuggothian
Joined: 11 years ago
Posts: 2581
13/04/2020 3:27 pm  

That site does fake cats too...

image

 

@djedi

What is FAS?


ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
13/04/2020 3:33 pm  

@christibrany

Fetal alcohol syndrome, I believe.


ReplyQuote
ignant666
(@ignant666)
Tangin
Joined: 14 years ago
Posts: 3110
13/04/2020 3:34 pm  

"Fetal alcohol syndrome", most likely.

[The Duck has beaten me to it, by a whole minute. Beaten by an anatidoid! The shame.]


ReplyQuote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
13/04/2020 8:01 pm  

Although we all know from our direct and unbiased experience with the infinite and eternal essence that there is no difference betwixt things, I beg to push the elevator button one floor lower so we might examine some differences in order to save the human race ...

ARTIFICIAL INTELLIGENCE

 

  • The ability of a computer or other machine to perform those activities that are normally thought to require intelligence.

  • The branch of computer science concerned with the development of machines having this ability.

  • Intelligence exhibited by an artificial (non-natural, man-made) entity.

When examining and anal-izing these silicon "entities," it is the word "intelligence" that requires a magified view under the Micro-scope (so named for Frater Elite).

1. I maintain that a silicon-digital system that performs (any) work as programmed is not intelligent, but merely a blind-force extension of the human mind. WYPIIWYG. What You Put In Is What You Get. This may be considered a "tool," a third circuit implement that hopefully saves time, money, and material (although, when operated, carbon particles are released into the atmo-sphere ("atom realm") that will eventually kill us, but let's overlook that long-term superstition for now).

2. Now if said silicon-digital (SD) system has memory banks with solutions (its own 777 with various courses of action), and said SD detects some "trigger," it can then refer to 777 for the correct action. If the criteria for selection have already been programmed, then we're still at level 1 (above). BUT, if the selection has criteria have NOT been pre-programmed, and the SD makes its selection based on criteria it has developed itself, then we have "true" intelligence.

Of course, if the intelligence is faulty, then we have an insane robot running around, even here in the forums.

So, for example, was the detection of Pneumonia in Wu-wei Wuhan simply a bell that went off due to detection of medical reports? If so, that was not AI; it was SD, doing its job.

Note: In terms of human reaction, it seems like any and all countries got notified way ahead of time, but delayed a lockdown until they had the Bad Angel running loose.

Anyway this AI has to be "I," able to think for itself, outside of the box that was originally programmed.

 

 


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
13/04/2020 8:14 pm  

Inspired by the AI discussion going on in this thread, I had a go with the Automatic Multispeaker Voice Cloning -Deep AI method described by Gilles Louppe in his 2019 Master Thesis.

 

(If someone is interested: there are no "ready" click-and-go programs available. The source code and files are available on Github https://github.com/CorentinJ/Real-Time-Voice-Cloning and the method is described: https://www.semanticscholar.org/paper/Master-thesis-%3A-Automatic-Multispeaker-Voice-Louppe/6529089a9e5af31c6221ca538193b36927dfcc83

 

I fed the audio from The Great Beast Speaks, about 1 minute of "The Pentagram" to the algorithm.

 

Then I made the Great Beast to recite some Lovecraft:

 

 

"That is not dead which can eternal lie. And with strange aeons even death may die."

 

This is the end result without "training" the deep learning algorithm at all. The best results would be possibly achieved by having someone create perhaps a victorian accent database with a wide variety of english words to choose from with their pronounciations and after that you would have to start "training" the algorithm to get the best possible end result. I personally don't have the time for that.

 

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


soz, christibrany, The HGA of a Duck and 1 people liked
ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
13/04/2020 8:19 pm  
Posted by: @shiva

BUT, if the selection has criteria have NOT been pre-programmed, and the SD makes its selection based on criteria it has developed itself, then we have "true" intelligence

Posted by: @shiva

Anyway this AI has to be "I," able to think for itself, outside of the box that was originally programmed.

I think its called "meta-programming" (the program can program itself). I think that's in the works too.

Folks like "RAW" would say humans could learn to program their own programming too. 🙂 


ReplyQuote
djedi
(@djedi)
𒇽𒅗𒅗𒈠
Joined: 9 months ago
Posts: 335
13/04/2020 9:34 pm  
Posted by: @behemoth

Then I made the Great Beast to recite some Lovecraft

Baleful is thy necromancy, or so it would seem. In truth, you have only conjured a Bardonian elementary clothed in the shells of the dead.

Joking aside, that's cool. Thanks for linking the files.

There's an old Harlan Ellison story called Laugh Track in which-

Spoiler
an Italian-American gentleman gets into show business and finds out his Aunt Babe's soul has been trapped on tape after her laugh was recorded on a sitcom laugh track. You may find it ironic to listen to the dead author recite it to you.

There's another Ellison story, lesser known, in which a computer becomes self-conscious and starts acting like the volcano god from the old testament. What was it called?


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
13/04/2020 9:58 pm  
Posted by: @djedi

Baleful is thy necromancy, or so it would seem. In truth, you have only conjured a Bardonian elementary clothed in the shells of the dead.

Joking aside, that's cool. Thanks for linking the files.

 

I think the Qliphothic part itself is the "sound profile" the program makes from the raw material you feed to it. It tries to mimic all sorts of possible sub-sounds of the human voice spectrum.

 

Here is for example the "base" sound for Crowley the program made for the profile before I used the command prompt to vocalize any sentences/words. This had some "reanimated dead" quality to it in my opinion. Call of the 31st Aethyr?

 

 

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
christibrany
(@christibrany)
Yuggothian
Joined: 11 years ago
Posts: 2581
13/04/2020 10:46 pm  
Posted by: @shiva

Note: In terms of human reaction, it seems like any and all countries got notified way ahead of time, but delayed a lockdown until they had the Bad Angel running loose.

Reminds me of every country and their astral counterpart telling the USA prior to 9/11 what and even when to expect and um nothing was done.  LIHOP (Let it happen on purpose) 

 

Never let a good LIHOP go to waste.  Especially with all the free coffee and high fructose 'maple' syrup. 


ReplyQuote
christibrany
(@christibrany)
Yuggothian
Joined: 11 years ago
Posts: 2581
13/04/2020 10:51 pm  

@behemoth

Yes doesn't sound English at all, more American. Interesting concept though for sure. 

I am tempted to turn your second 'base' sound into a (creepy of course) song for posting on my soundcloud (plug https://soundcloud.com/chris-stibrany )


ReplyQuote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
13/04/2020 10:58 pm  
Posted by: @duck

Folks like "RAW" would say humans could learn to program their own programming too.

The 6th, Electric, Metaprogramming circuit is located at the Geburah B&B, open to any and all 6=5s.

 


ReplyQuote
 Anonymous
Joined: 51 years ago
Posts: 0
13/04/2020 11:00 pm  
 
A holographic AI Crowley leading ceremonies at Boleskine would be amazing!
 
Posted by: @duck

humans could learn to program their own programming

That's the basic idea to spiritual/personal development. Improving those inner qualities. 😉

It's plainly spoken of by P.D. Ouspensky in The Psychology of Man's Possible Evolution.

Our fundamental idea shall be that man as we know him is not a completed being; that nature develops him only up to a certain point and then leaves him, either to develop further, by his own efforts and devices, or to live and die such as he was born, or to degenerate and lose capacity for development. Evolution of man in this case will mean the development of certain inner qualities and features which usually remain undeveloped, and cannot develop by themselves.

This work is said to be done to overcome several obstacles. An important one is set out in the introduction to the above book. 

I found that the chief difficulty for most people was to realise that they had really heard new things, that is, things that they had never heard before. They did not formulate it for themselves, but in fact they always tried to contradict this in their minds and translate what they heard into their habitual language, whatever it happened to be.

 


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
14/04/2020 5:18 am  
Posted by: @christibrany

Yes doesn't sound English at all, more American. Interesting concept though for sure. 

I am tempted to turn your second 'base' sound into a (creepy of course) song for posting on my soundcloud (plug https://soundcloud.com/chris-stibrany )

 

 

Yes go ahead, I really cannot make any copyright claims to a cloned voice. (I see you're a fan of Coil, huh? great band...)

 

I am pretty sure that the big government already has extremely advanced voice cloning -techniques, but they are not available to public (mis)use. Usually it is a situation where DARPA rolls out their 20+ year old prototypes and pretends it’s the “state of the art” on YouTube/TV, imagine that 20+ years down the road, at the level of advancements made by the hour these days...

 

There are apps available like LyreBird (closed at the moment) and some others for voice cloning, but you have to fill out forms and specify the reasons why are you cloning this voice and for what use.

 

The misuse potential is of course identity theft that some criminals might use. Or someone cloning the voice of officials/celebrities saying something controversial. These audio "deepfakes" are still extremely new thing (compared to video deepfakes), having been around only for like few years at most ( for public use ).

 

Posted by: @xon

A holographic AI Crowley leading ceremonies at Boleskine would be amazing!

 

I think there will be lots of monetizing on this technology (already is). Buying Stephen King -voice profile to listen Stephen King audiobooks with the narration done by Stephen King DeepFake etc.

 

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
14/04/2020 10:56 am  
Posted by: @behemoth

Inspired by the AI discussion going on in this thread, I had a go with the Automatic Multispeaker Voice Cloning -Deep AI method described by Gilles Louppe in his 2019 Master Thesis.

 

(If someone is interested: there are no "ready" click-and-go programs available. The source code and files are available on Github https://github.com/CorentinJ/Real-Time-Voice-Cloning and the method is described: https://www.semanticscholar.org/paper/Master-thesis-%3A-Automatic-Multispeaker-Voice-Louppe/6529089a9e5af31c6221ca538193b36927dfcc83

 

I fed the audio from The Great Beast Speaks, about 1 minute of "The Pentagram" to the algorithm.

 

Then I made the Great Beast to recite some Lovecraft:

 

 

 

 

 

Some little follow-up on this voice replication if someone wants to try the software on their own:

 

The problem with replicating the voice of Crowley is that we only have "The Great Beast Speaks" collection that is composed only of early 20th century recordings of Crowley's voice.

 

The recording devices back then could only capture small part of the voice frequency spectrum and and the recordings itself are not the best quality and fail itself to reproduce the whole spectrum.

 

Then you feed the program so it has to do lots of "guesswork" and you end up with a clear digital signal that sounds extremely unnatural.

 

To make it seem more credible: it is best to add some sort of saturator or other VST effects. I for example used the Vintager Toy VST with "1935" preset and the end result is much more akin to the recordings of Crowley where the sound was cloned from to begin with:

saturator

 

After the "1935 Vintager"-preset:

 

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
 Anonymous
Joined: 51 years ago
Posts: 0
14/04/2020 5:06 pm  

While the data the program was given was very limited the result was great. That the input was poetic would help with the given task of reciting that line from Lovecraft. Still, the cadence is wonderful. As public releases improve it will not only be great for AI performances but a powerful tool for live actors to draw from and, in some specific circumstances, for dramatic ceremonialists who want to mimic that style.


ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
14/04/2020 7:13 pm  

I thought I should give an update on this (and a potential warning)

This was brought to my attention by the /x/ - Paranormal board on the infamous 4chan (there's good info to be found anywhere if its your will to do so). The subject of the thread there was that though this A.I. produces realistic faces when its working properly but when the algorithm doesn't know what to do (eg putting people in the background of the image) it produces some surreal and sometimes almost terrifying results. A few examples I found:

creepy01
creepy02
creepy03

 

This one was teeth where the eye should be:

creepy04

 

As you can see the results are quite bizarre, even "Hieronymous Bosch-ian" or "Qliphotic".

 

After spending a lot of time browsing these pics (it was fun), I had some bizarre dreams and semi-dreams just before sleeping. These could be described as "Qliphotic". I warn people from spending too much time on this AI site, even if the faces you see look normal they may have something in them that has creepy a sub-conscious effect. If you are going to spend time on that site (it is quite fun) I recommend you perform your banishing ritual before doings so and/or praying to your god.


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
14/04/2020 7:34 pm  
Posted by: @xon

While the data the program was given was very limited the result was great. That the input was poetic would help with the given task of reciting that line from Lovecraft. Still, the cadence is wonderful. As public releases improve it will not only be great for AI performances but a powerful tool for live actors to draw from and, in some specific circumstances, for dramatic ceremonialists who want to mimic that style.

 

Another problem arising to clone Crowley's voice is that I do not believe we have a recorded demonstration of his normal, daily speech voice that Crowley would have used in a normal conversation.

 

All the tracks on The Great Beast Speaks are performances and poetic recitations. Some of the stuff is not even spoken in English, but in Enochian.

 

Most of the tracks on The Great Beast Speaks are also "melodic" and have invocation -like quality to them, rather than a narrative reading or formal presentation style and this particular deep learning program is/was designed for a simple audiobook style text to voice conversion.

 

So all of the words and phrases created have a "declaration" -like quality to them.

 

Example:

"The word of the Law is Thelema"

 

The same kind of problem would arise if you tried to create a voice profile of Adolf Hitler using mouth foaming nazi rallies as your only audiofeed source.

(Before they found the secretly recorded Mannerheim-Hitler dinner tapes that contained private conversational tone of the Führer)

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
djedi
(@djedi)
𒇽𒅗𒅗𒈠
Joined: 9 months ago
Posts: 335
14/04/2020 7:54 pm  
Posted by: @duck

/x/ - Paranormal board on the infamous 4chan

Before 8chan got taken down for El Paso, there was a small community of ariosophists and 'esoteric hitlerists' on its /x/ board who were a hoot to talk to, always trying to put together rituals to fight off the collective psyche of Zion or whatever other Frankenstein's monster du jour they had sewn up. I left a little before the end, though, after a feud with the pedophile who they contracted out to rebuild the website, so I don't know what became of all those wonderfully strange people. 8kun isn't the same, I reckon because it doesn't have a /pol/ or any political board comparable to 8chan's, but it does have its own /x/ that you might find more interesting than 4chan's.

You would think that, with more people using the internet, it would become a weirder and more interesting place. I've found it to be the opposite.


ReplyQuote
RuneLogIX
(@runelogix)
Magister
Joined: 15 years ago
Posts: 369
14/04/2020 9:04 pm  

@djedi chan culture has been on the decline for awhile now. it arguably peaked between 2011-2016. I also rediscovered that 8kun is back sort of. However with caturday taking over twitter and qanon making waves over brainlets its influence seems more diffuse (dare I say, mainstream?) and unidentifiable. the pranks that characterized it are largely done with and urban legends now, in its place is Tiktok dances and eceleb's, RIP internet sewer of yesteryear!

Force and Fire is not metaphorical. In Prophetes Veritas Venit.


djedi liked
ReplyQuote
 Anonymous
Joined: 51 years ago
Posts: 0
15/04/2020 5:02 pm  
Posted by: @behemoth

Most of the tracks on The Great Beast Speaks are also "melodic" and have invocation -like quality to them, rather than a narrative reading or formal presentation style and this particular deep learning program is/was designed for a simple audiobook style text to voice conversion.

Which, thanks to your work with the program and posts, shows that that melodic style can be input and the program will return a ball-park conversion in that style.

Would love to hear how it does with the opening line to Resh, if you're willing:

Hail unto Thee who art Ra in Thy rising, even unto Thee who art Ra in Thy strength...

Actually, after listening to those others I can kind of imagine how it sounds now.

 

Posted by: @duck
 

Hieronymous Bosch-ian

Not sure it's that talented.

I got one of those you mentioned

AI face of child 2

 

There were some other artifacts that showed up, like random blue streaks. Perhaps, it's trying for braids or not scanning out headwear. Reminds of the plot to Kingsman 2(don't partake around your kids!).

image

 

The first image I got was cute though. It looks like someone is waving in the reflection of the eyes although the background of the reflection appears to be indoors and doesn't match the background of the pic which the image of the child looks past onto.  There's also some obvious roughness around the chin and scarf.

AI face of child

 

This one was attractive although has some mismatching of skin tone around the jawline towards the ears and around the side of the face.

attractive woman

 


ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
15/04/2020 7:15 pm  
Posted by: @xon
Posted by: @behemoth

Most of the tracks on The Great Beast Speaks are also "melodic" and have invocation -like quality to them, rather than a narrative reading or formal presentation style and this particular deep learning program is/was designed for a simple audiobook style text to voice conversion.

Which, thanks to your work with the program and posts, shows that that melodic style can be input and the program will return a ball-park conversion in that style.

Would love to hear how it does with the opening line to Resh, if you're willing:

Hail unto Thee who art Ra in Thy rising, even unto Thee who art Ra in Thy strength...

Actually, after listening to those others I can kind of imagine how it sounds now.

 

 

Sadly I have run out of CUDA memory this program uses for the Deep Neural Network library and when I tried to open the program moments ago I could not even get it up and running. It was just a whim of the moment on my not-so-great laptop.

 

Most of my better computer hardware is at the university and I cannot get to it due to COVID lockdowns.

 

I might get some of the stuff during the summer months and try it again, but for the moment the Shells have been banished unto their abodes and habitations and we have entered into a period of Silence.

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


You liked
ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
16/04/2020 3:06 am  

@behemoth

Thanks for all the input you have provided. I put this thread in the "Stuff" section as I couldn't see any connection to AC, You have found a significant connection with him.

The discussion here has been thought provoking, bear with me while I try to piece some of this together without rambling on too much:

Posted by: @behemoth

there is certain hollowness to the sound with computer software especially when it comes to generating or mimicking the human voice and analog sound in general

Yes, it still has some way to go before it is entirely convincing. Out of the methods of simulating a real musical instrument, so far I have been the most impressed with Physical Modelling.

A short example with strings:

https://www.youtube.com/watch?v=iTu7BeLd3Y8

 

Think of how a wav file of say a violin note can be produced. One way is to record the violin with a microphone. You can then see the waveform on the computer. As it now exists in digital form it is possible in theory to produce the same waveform inside the computer alone. This could be done with physical modelling where calculations are made of how an actual sound wave behaves in the strings and resonating chamber of a violin. With enough calculations the resulting waveform would be identical to the recorded one. I'm going somewhere with all this, bear with me. 🙂 

 

The article about the recreated larynx of an Egyptian priest:

https://www.bbc.co.uk/news/world-middle-east-51223828

 

It mentions how the scientists used 3d printing to achieve this. Seeing as they used measurements and calculations but applied them to make a physical recreation, the same measuments and calculations could be used on a computer simulation alone. This would be an example of "Physical Modelling". Now, I believe A.C's body was cremated so none of his remains can be measured and this data used to support a "physical model". What we do have are several photos of him from various angles and possibly some height/weight measurements. These could be used to "interpolate" the shape of his skull for example. There are also the yoga photos of him nude so the shape of his body could also be used in this model. The skull shape and body shape affect the resonance of the sound. A larynx would be simulated and the recordings of AC's voice could be used to determine this.

Recreating his voice would then be a "2 pronged attack" where the recordings of his voice would reinforce the physical model and vice versa.

Additional data could be acquired by using a talented vocal impressionist to listen to the originals and mimic them.

 

A.I. training sets could eventually be used for just about anything measurable. All of A.C's writings could be used along with all biographies to develop an "artificial personality". A training set for accents is another possibility.


Behemoth liked
ReplyQuote
Behemoth
(@behemoth)
Member
Joined: 8 months ago
Posts: 74
16/04/2020 6:10 am  
Posted by: @duck

Yes, it still has some way to go before it is entirely convincing. Out of the methods of simulating a real musical instrument, so far I have been the most impressed with Physical Modelling.

A short example with strings:

https://www.youtube.com/watch?v=iTu7BeLd3Y8

 

Even though that particular video does sound impressive: one should approach with a healthy dose of skepticism. It seems like an advertisement for the "Arché Collection" library and this particular promo video might have so much audio wizardry and mastering going on that it has nothing to do with the real base sound of the library.

 

Most classical composers, who want to use "artificial" instruments, usually buy something similar to VSL https://en.wikipedia.org/wiki/Vienna_Symphonic_Library (extremely expensive stuff) and run the libraries through different sampler programs and DAWs.

 

To make a price comparison: that particular Arché Collection sells for 120 bucks a month or $49 for each instrument while the Vienna Sound Libraries collections cost over $7000. If you buy the whole VSL libraries you probably end with a bill of $ € 30000 to 50000. But price tags are not the only indicator in terms of music production: a good professional with decades of experience can even work with 90s ROMpler keyboards and fool the untrained ear.

 

Posted by: @duck

This could be done with physical modelling where calculations are made of how an actual sound wave behaves in the strings and resonating chamber of a violin. With enough calculations the resulting waveform would be identical to the recorded one. I'm going somewhere with all this, bear with me

 

This is possible / has been done for years. We can reproduce the sound/waveform of a violin note almost 1:1 identically and use different impulse response libraries for the sound https://en.wikipedia.org/wiki/Impulse_response#Acoustic_and_audio_applications

 

Different impulse responses have been captured inside cathedrals, music concert halls, studios etc. and they already are "good enough". The direct signal is driven through the impulse response modulator to achieve the acoustic effect of said place. Example:

https://www.youtube.com/watch?v=U2fOqSFguU0

 

Posted by: @duck

The article about the recreated larynx of an Egyptian priest:

 

https://www.bbc.co.uk/news/world-middle-east-51223828

 

This would be an example of "Physical Modelling". Now, I believe A.C's body was cremated so none of his remains can be measured and this data used to support a "physical model". What we do have are several photos of him from various angles and possibly some height/weight measurements. These could be used to "interpolate" the shape of his skull for example. There are also the yoga photos of him nude so the shape of his body could also be used in this model. The skull shape and body shape affect the resonance of the sound. A larynx would be simulated and the recordings of AC's voice could be used to determine this.

 

Even if you could get the larynx itself there is still so much else going on that contributes to the unique characteristics of a person's voice. Crowley's (pipe) smoking habit comes to mind. Also other factors infinite & unknown like asthma may have an effect. Most opiate abusers I have met also have a bit of a raspy, nasal voice.

Behold now Behemoth, which I made with thee; he eateth grass as an ox.


ReplyQuote
The HGA of a Duck
(@duck)
Member
Joined: 7 months ago
Posts: 624
17/04/2020 6:39 am  
Posted by: @shiva

Tetrahedral coordination is a major structural motif in silicon chemistry just as it is for carbon chemistry

The science is a bit beyond me but I believe it is referring to the shapes that atoms and molecules naturally form into.

https://en.wikipedia.org/wiki/Tetrahedral_molecular_geometry

 

One of these shapes is the Tetrahedron. Tetraheral numbers are those those that can be represented by spheres stacked up to form the shape:

 

One of these tetrahedral numbers is 10, seen in the animation as the first 3 layers of spheres. This kind of visualisation would allow for a 3-dimensional representation of the Tree of Life for example.

Another one of these numbers is the number of verses in the Book of the Law, 220:

tetra 220

Each one of the verses could be put in its own little sphere and then they could all be stacked up.

I find it quite "satisfying" that AL makes such a nice shape and believe it could be significant.


ReplyQuote
Shiva
(@shiva)
Not a Rajah
Joined: 13 years ago
Posts: 4944
17/04/2020 8:14 pm  
Posted by: @duck

The science is a bit beyond me but I believe it is referring to the shapes that atoms and molecules naturally form into.

Yes. We carbon based CHONs (Carbon, Hydogen, Oxygen, Nitrogen) live in a world based on SIX. Snowflakes, sunrises, etc all radiate six rays. The so-called "inorganic beings" radiate five rays (in their molecular structure), yet five-ray radiation is virtually (completely?) unknown in our world.

There are technical terms and straight-line references to support this [^] statement.

But I can't remember, due to senility or simply not caring to memorize the terms because there's nothing I can do about it right now. But I will look these little gems up, and I will return to overload overshadow our first circuits with fear ... or maybe just mild amusement.

 


ReplyQuote
christibrany
(@christibrany)
Yuggothian
Joined: 11 years ago
Posts: 2581
17/04/2020 10:54 pm  
Posted by: @behemoth

Most classical composers, who want to use "artificial" instruments, usually buy something similar to VSL https://en.wikipedia.org/wiki/Vienna_Symphonic_Library (extremely expensive stuff) and run the libraries through different sampler programs and DAWs.

 

I am not a professional composer so I dont have the budget, but what I do use and am a firm believer of is sound fonts.  .sfz if I recall correctly.  I write some classical stuff sometimes that I post online and most of the string are sound fonts.   Came from real instruments. 


ReplyQuote
Share: