larme

For me creating the exact sound is not very interesting from sound designing perspective. You can always sample the real instrument.

Like physical modeling synthesis, the interesting part is to compress the sound to some parameters that you can tweak and generate new sounds

Another approach is VAE, which also you give your some latent embedding, you can tweak the embedding to generate new sound. However the meaning of this embedding is not explicit.

People who are interested in this application should check synplant[0]. It has a ML technology called "Genopatch" which gives you 2 functionality:

1. you can try to describe a sound with some tags and it will try to generate a sound to capture the feeling of these tags

2.you can feed it with a sound sample and it will try to re-synthesize the sound with its synth engine. Though the end result will usually be just a "re-imagined" version of your input sample.

My guess is the underlying model is not a "deep" model. The main benefit is that the end result is not a wave file, but a list of generated parameters that can be synthesized by the synthplant engine. And now it comes the interesting part: you can tweak these parameters to finetune the generated sound. These parameters have actual meanings (FM ratio, reverb etc.)

[0] https://soniccharge.com/synplant

Before LLM there has already been Fields medalist[0] who creates professional software[1].

[0] https://en.wikipedia.org/wiki/Martin_Hairer

[1] https://www.hairersoft.com/

This description sounds like a cult.

It's a big democracy red flag when a majority wants to take a lot from a tiny minority

It’s not “taking”. The rich give out some money so the society has a higher probability to stay peaceful. or a violent revolution may happen.

This is really a win win situation

Go is not high performance enough. Like what others said, you implement the high performance part in C++ and use python to glue them.

Try pair reading it with a good thinking LLM like GPT 5.5 or Claude Opus. I found it help me a lot.

I have started learning SBCL internal from the beginning of this year with the help of GPT, and I really want to contribute to SBCL compiler someday in future.

Some of these models are open weight. You can try hosting them and do the price calculation yourself.

They also publish papers talking about how to save kv cache and computation powers. Because currently they don't have the most powerful nvidia cards, training and inference efficiency is very import for them.

That's amazing. I wonder how accurate is the synth engine?

It’s true that American companies have benefited from sending jobs overseas. Instead of trying to stop them, perhaps we could explore ways to share this money more fairly across the country.

Have you tried pocket 386? https://arstechnica.com/gadgets/2024/08/a-few-weeks-with-the...

Thanks for the answers!

I'm very interesting in this field (realtime audio + GPU programming). How do you deal with the latency? Do you send or multiple single vectors/buffers to GPU?

Also I think because samples in one channel need to be processed sequentially, does that mean mono audio processing won't benefit a lot from GPU programming. Or maybe you are dealing with spectral signal processing?

Wow that's really thorough

What features has Google killed, do you think? The problem is more like google can implement whatever feature they want and force it into web standard

How often has this been a problem, do you think? very often

What other browsers have gone ahead, with Google holding out? Google is holding out in a sense that no other people can implement a feature-complete browser. Google is killing the "open standard" web by make the standard impossible.

I use rpi zero 2 as a coding server (paired with my iphone and a splitted keyboard) when I have slow or no internet. It's quite usable if you just mosh/ssh into it and use emacs/vim to code.

The same logic applies to capital punishment

It indicates how much does google search suck nowadays.

Ye, SE Lain has lots of tech reference. This website lists many of them: https://www.cjas.org/~leng/lain.htm

entrepreneurs suck anyway.

I didn't use google's image search for a long time so I don't know that. But I think kagi pays to use google's data as one of their source. So it's possible the image search results are the same. However the different order/ranking is what separated kagi and google.

Try jmp.chat, you can use an XMPP client to receive phone calls and sms.

I use the following services:

- phone number: jmp.chat and a textnow number as backup

- email: fastmail

- search: kagi

- map: apple map

don't act like kids calling their parents to stop an argument

try combining a type c male to USB female adapter and a USB A to type c cable together to get a working type c to type c cable

I use a docker image as my whole development environment and the experience is similar. I can do whatever I want inside the container, and all will gone after I restart it. For the changes I like it, I edit the dockerfile and build a new image. When I get a new machine, I just install docker and pull the image and start coding.

Not the same but there's a netlabel called 20kbps[0]. Most of their albums are mp3s around 10-20 MB in total.

[0] https://20kbps.net/index2.htm

Lockfile does not work perfectly for me because it does not lock python version. Using a lock file with different python version or even same version but different OS may fail.

I'd like to add that most of these text only talking about inference part. This book (I also purchased the draft version) has training and finetuning in the TOC. I assume it will include materials about how to do training and finetuning from scratch.

https://jaykmody.com/blog/gpt-from-scratch/ for a gpt2 inference engine in numpy

then

https://www.dipkumar.dev/becoming-the-unbeatable/posts/gpt-k... for adding a kv cache implementation

It's just a fucking book. Don't push your shits like js or SPA or d3 or webgpu to a fucking book. I just want to read it like a dead tree book.