 So, there's some acapella singing in this one. There's buckets of pitch correction, in a very unsubtle way. To the point where it stylistically changes the performance. Makes me wonder. Do most people hear it? Do they expect it? Is hyper-realistic singing now our baseline for what regular singing is supposed to sound like? A blog post on the Kelly Criterion got posted on HN today. This is the same Kelly of the Kelly-Lochbaum scattering junction, which was part of the first digital simulation of the human vocal tract in the 60s at bell labs, which also powered the first singing computer. I get the impression that Bell Labs made everyone work on speech R+D at some point. "Sound Mantra" (2023) is an audiovisual etude series exploring circular phrases, sound objects, minimalist notation, and time manipulation.  @paul I'm hooked after listening to some of these on loop. How can I get your stack running? It seems like the pieces are: sndkit -> mnolth -> SoundMantra Am I missing anything? I can't tell how to get sndkit from https://pbat.ch/wiki/sndkit. The other pieces I see the 'source' links for on their respective pages. Can anyone help me find pokemon names that sound like speech physiology terms? This is important. Randomly discovered the "Donny Most" Family Gag gag this morning, which features a surprisingly pretty 4 part singing chorale. It was composed with lyrics and performed with real performers just for this 1 minute bit. What's funny to me is that "Donny" is very close to the latin word "Dona", as in "Dona Nobis Pacem" from the Agnus Dei, which is word you hear all the time in sacred choral music. Distilled thoughts on organized sound over the past year: Physiology -> Mechanism -> Vocabulary -> Physiology Of course ADC would have a presentation whose whole topic is about explaining to me why I shouldn't do things. Week 8 at RC: singing drawing tablets, hacking the tic80 sound engine, NANDs, and more! https://pbat.ch/recurse/week08/ Poke: procedurally generated tiny critter sounds using physical models: https://pbat.ch/recurse/demos/poke/ @patchlore my cats were convinced enough to go running to the back window to see what was going on I've been thinking about maybe trying to do an arrangement of "Milowda" from the Expanse soundtrack. Found a transcription, did an initial chord sketch to hear what it would sound like before committing more time into it.  @patchlore (Weird seeing someone else using my avatar. Not complaining, you’re welcome to it, it just provoked a moment of confusion. “Wait, did I post that?”) The synthetic singing ensemble sings a progression now. Also tried to work on the mix a little bit. Getting synthetic voices to blend is difficult! I have a new found respect for the real thing. Added a quick hack to adjust vocal tract length to my singing synthesizer prototype [0]. I really didn't think it would work as well as it does. There's no smoothing yet, so adjusting it introduces little clicks. Actually, the sound reminds me of those pull/stretch tube toys. From a physics perspective, the model is actually quite similar to that!  @patchlore You've probably seen this, but: https://dood.al/pinktrombone/ I've always been curious how things like this work. If you decide to do a writeup, please publicize 🙂  @patchlore Those clicks don't sound like digital artefacts. My brain is treating this like a physical object, and trying to locate the source of the clicking. (I think they're coming from the far end of the tube, where there's some kind of slip-and-stick motion.) The pitch of this singer is controlled using a technique I invented called Gesture Synthesis. I have made an interactive demo for it [0]: Hit Begin. Singing should play. (Think of pitch changes in the melody as a smooth line.) Change the Tempo slider. The pitch "line" smoothly expands and contracts with the tempo changes. Almost instantly. This is Gesture Synthesis. I'm writing a vocal synthesizer in Rust! Some initial scaffolding: https://github.com/paulbatchelor/voxbox All it has is a very simple WAV writer for mono 44.1khz 16-bit audio. Offline rendering to disk is very helpful for prototyping and debugging. Next up: porting a glottal source model. The one I use is based on the LF model with pulsed glottal noise for breathiness. Original C code here: https://git.sr.ht/~pbatch/mnodes/tree/master/item/glot/glot.c I am logging my time and knowledge acquisition during my time at the Recurse Center Summer 1 Batch! Check it out here: https://github.com/PaulBatchelor/Recurse |
Gregory's Server
@patchlore You might find this interesting: https://insamjournal.com/index.php/ij/article/download/97/th-8/300
elicit.com summary:
The proliferation of audio manipulation in pop music can negatively impact young learners' confidence and vocal health by presenting unachievable performances.