Submind YouTube summaries
Thumbnail for Interactive Phonetic Simulator: Biomechanics of Speech Synthesis

Interactive Phonetic Simulator: Biomechanics of Speech Synthesis

Watch on YouTube

Video summary

The video introduces an innovative interactive tool called the Phonetic Simulator, which visualizes the complex biomechanics of human speech synthesis in real-time. This application allows users to manually control and observe the movement of anatomical structures such as the tongue, jaw, teeth, and lips, effectively turning the screen into a flight simulator for piloting a human vocal tract. By manipulating sliders that adjust parameters like jaw depression and tongue height, viewers can witness how physical changes in the oral cavity directly alter the generated sound, bridging the gap between abstract audio output and tangible biological motion. At the core of this simulation is an explanation of acoustic filtering theory, where the vocal tract acts as a resonant chamber similar to a flute rather than a vibrating string instrument. The tool demonstrates how specific sounds are produced by changing the volume and shape of air cavities; for instance, raising the tongue lowers the first formant while pushing it forward affects the second formant. The interface includes an integrated IPA vowel quadrilateral that maps these movements geometrically, tracking the highest point of the tongue to show exactly which sound is being articulated based on the precise positioning of the anatomy within the simulated throat and mouth. Beyond passive observation, the simulator offers active engagement by allowing users to type custom sentences or manually drag the pink tongue body to perform articulatory gymnastics. The system breaks down complex words like "bionic chaos" into their constituent phonetic parts, illustrating events such as bilabial bursts for 'b' and velopharyngeal port openings for nasal sounds. While the current version focuses primarily on visualizing the kinematics of the mouth and tongue rather than perfecting synthetic audio quality, it successfully uses built-in browser text-to-speech to play back sentences, enabling users to hear how their manual adjustments influence the resulting acoustic output. The video concludes by pondering the future implications of this technology, suggesting that digital voices may soon evolve from simple audio files into entities literally "spoken" by virtual lungs and vocal tracts. By simulating the precise physics of biological speech production, the tool provides a unique educational platform for understanding how a combination of physical cavities and aerodynamic turbulence generates the entire spectrum of human language. The developers invite feedback on improving the audio clarity and pronunciation algorithms, emphasizing that this project is an ongoing experiment to better visualize the intricate mechanics behind every word we speak.
Read the full video transcript
Okay, we have this new tool on the website. It's a bio acoustic articulatory phonetics and vocal track animatics. It's a bit of a muff. Could have came up with a better better name. Well, the better name is phonetic simulator. So, you can see we can control this mouth and tongue. I don't know if it's clear that this is mouth. So the teeth, the tongue, the jaw should be pretty clear. If you have any ideas how to improve this, uh do let me know. You can control it from the from the control panel as well. You can pronounce certain words like bionic chaos. It's not super clean, but this is just um essentially building a a text to speech from scratch. Yeah. When you have the sound on, it keeps vibrating the road. But also when you change the when you change the position of the tongue or the jaw, the mouth open and close, it changes the sound as your mouth does. and we will do an overview by notebook and lamb and let me know if you have any questions or comments on it. >> Right now there is a tornado of pressurized air vibrating inside your throat, bouncing off the wet walls of your mouth and escaping your lips as language. We rarely think about it, but you are constantly operating a highly complex biological wind instrument. Welcome to this deep dive. Today, our mission is to act as your audio guide while you run the demo mode of the English speech articulator and phonetics visualizer over at bioniccast.com, pronounced bionic chaos. >> Yeah, we're basically decoding the fascinating biomechanics of human speech right alongside the simulation you're watching. >> Okay, let's unpack this. If you look at the main visualizer canvas, it's well, it's pretty mesmerizing. You've got this pink tongue body moving in real time, >> right? And you can see those blue formant monitors tracking the frequencies plus that uh labial view showing the actual lip shapes. >> Exactly. And this interface feels exactly like a flight simulator but you know instead of flying a Boeing you are manually piloting a human mouth. >> That is like the perfect analogy. And that grid in the upper right the integrated IPA vow quadrilateral. It's basically a literal map of your oral cavity. >> Wait really I mean looking at it just looks like a geometric map. >> Well it is. The top edge is the roof of your mouth. The bottom is your jaw dropped. The left side represents your front teeth. And the right side is your throat. >> Oh, so when that yellow IPA cursor moves around, it's literally just tracking where the highest point of your tongue is. >> Yeah, you got it. It's tracking exactly what the anatomy is doing to create a specific sound. >> Okay. But if the canvas is the windshield, the sliders on the right are the cockpit controls, right? Like jaw depression and tongue dorsome height. >> Exactly. and tongue backness too. Moving those is what actually changes the physical shape of the mouth. >> But how do these physical movements actually create the sounds? Is it like pressing frets on a guitar string? >> Well, it's actually much closer to blowing into a flute. You aren't shortening a vibrating string. You're changing the volume of a resonant chamber. >> Oh, right. This goes back to Gunnar Fance acoustic theory of speech production, doesn't it? >> Yeah, exactly. Speech is really just acoustic filtering applied to aerodynamic turbulence. Your glauus is the source just vibrating away in your throat. >> And then that 17.5 cm vocal tract acts as a complex acoustic tube acting as the filter. >> Exactly. When you move those sliders, you're physically squeezing and expanding different sections of that tube. >> Here's where it gets really interesting. Wait, so the mysterious formance like F1 and F2, they aren't just abstract math. F1 is directly controlled by how high my tongue is. >> Yep. They are intensely physical. And F2 is just well how far back it sits in your mouth. >> Oh wow. So it's like blowing over a glass bottle, right? A smaller volume of air means a higher pitch. >> You nailed the physics. Dropping your jaw opens the front cavity which raises F1. Pushing your tongue forward shrinks that front cavity raising F2. >> Okay, so let's connect that physical theory to the demo mode you're watching where the graphine to foam engine orchestrates these exact movements. >> Yeah, it translates text into precise physical kinematics. It's wild to watch. >> Watch how the tool breaks down the custom word bionic chaos. You see the complete kinematic sequence like the bilabial burst for the bee. Right at the start. >> Yeah. The lips slam completely shut. And then uh pay attention to the N where the veloperial port opens. >> Right. Which routes the acoustic energy right through the nose. >> Exactly. And then for the S, you get that high velocity airream directly across the incizers. What's fascinating here is well, you don't just have to watch the demo run. >> No, totally. You can actively click and drag the pink tongue body manually or type custom sentences to watch these complex articuly gymnastics unfold in real time. >> We've basically just explored how a simple combination of physical cavities and a 17.5 cm acoustic tube generates the entire spectrum of human language. >> It really is incredible when you break down the actual physics of it all. >> It really is. So leave you with this final thought to ponder while you keep playing with the demo. If an interactive module can perfectly synthesize speech by simulating the precise physics of physical lips, tongues, and pallets. >> Yeah, literally simulating the biology, >> right? How long until the digital voices we interact with daily aren't just audio files, but are literally being spoken by a virtual set of digital lungs? >> Okay, this was not a lamb overview. 4 minutes, 4 and a half minutes. I'm not sure if it did well on showcasing the tool in terms of the especially in terms of the sound the synthetic sound activation. So playing with our sales. So this is it. I guess that's so that's it pronouncing bionic chaos if that wasn't clear. There is a description at the bottom of the page. Let's take this uh paragraph. Pop it. If you pop it into this text box, it will pronounce Okay, that was difficult to understand. So, currently it's just focusing on how the tongue and the mouth open, close, the position of the mouth and the tongue. Um, we will we could try improve the pronunciation. >> Yeah, it's not stopping the sound because the throat keeps vibrating for sound generation. Okay. So, you get the picture and streaming the sentence will uh essentially in this browser uh there's no uh there's no text to speech built in into the browser. If you're using Chrome or in our browser that has a text to speech, uh this uh stream sentence button will activate that will repeat whatever is in this text box using your inbuilt uh uh text to speech. Yes, I think you get the picture of what it does. It's available. Go check it out on bony.com/fonetics seam. It's for phonetics simulator. It's also on the landing page and we have many many other tools on the landing page as well. So go check them out as well and I'll see you next time. Bye-bye. trying to wave to like a you can actually activate it with a keyboard or press the buttons here. Yeah. So, check out this EMG simulator as well, but