This website likely uses a semantic vector space like CLIP or BERT space. Which is based on co-occurrence of words, it doesn't have a harder logic behind it. Soviets probably had cylinders in some machinery which is why it pulled.
i suppose this is an environment noise generator, with duck calls just being recordings. i wish you could control ducks individually, after i threw peas they just hang out near mic forever
It does appear to synthesize the audio with a wasm module compiled from rust, presumably developed as a python script first based on the source comments.
To the credit of the author there is also an attribution file which describes a process of deriving the models from recordings (none of the techniques are very novel, but it's fairly impressive to see automated -- edit: I assume the project is LLM-generated) https://github.com/celphase/duck-pond/blob/main/ATTRIBUTION....
The project was created with significant LLM use, I'd add claude as a co-contributor if I could do so manually. I don't personally like to have it added as a co-author on commits on my own projects, though I always make sure it's there if I submit PRs using it.
Honestly I can't take credit for any major input on it other than a few technical choices. Audio is entirely outside my wheelhouse beyond what happens at the platform API level.
The process involved running the LLM in a loop to gradually correct parameters of a synthesis stack to have the spectrogram of the output match the spectrogram of a real recorded sample. One funny part is where the synthesis was marked by a classification model as more duck than the real duck sample.
I'm super late to this, but I'm looking through the code, being inspired to figure out how it works, and one thing you mention is that synth.rs is a port of synth.py. Is the python code anywhere I would LOVE to get hold of it, rather than make a dirty port from scratch if possible.
> the synthesis was marked by a classification model as more duck than the real duck sample
The quality of a recording can throw off most classifiers.
Models we have today are so dependent on the quality of training data, that it's questionable how we will sustain long term what we are currently calling "AI".
i.e. What happens when the cameras and microphones everyone has in their pockets are significantly higher quality and the AI model hasn't had a total refresh in a long time? How does that cause classification to drift enough to fail?
This actually uses no recordings. All duck sounds are modeled live, with the quacks and rhaebs specifically modeling the duck's syrinx and vocal tract. The github repository has a full technical breakdown.
Some exotic fish and frog species are completely transparent. So one might wonder what makes meat opaque. Then wonder how to get the opaque out of other meat. I don't know where doritos comes in.
They use tartrazine that is a common dye for food, not crushed Doritos. They need a lot of dye to change the internal refraction index and make the meat transparent. So they choose it. From the research article https://pmc.ncbi.nlm.nih.gov/articles/PMC11931656/
> It is a hygroscopic compound with a high solubility in water and minimal reactivity. The European Food Safety Authority (EFSA) has concluded that this molecule exhibits no adverse effects or systemic toxicity at doses up to 2 g per kilogram of body weight (14). As a result, it is widely used as an artificial food dye in products such as flavored chips (for example, Doritos).
Vertebrate blood is one thing that is opaque due to hemoglobin, and it cannot be made transparent. The same is true for the pigmented retina of the eyes. So transparent frogs and fishes still have opaque eyes and blood vessels.
Where pigments like the heme of the blood or the melanin of the skin or the visual pigments of the retina are absent, the opacity is due only to the variable refractive index, which causes light scattering, and in this case the tissue can be made transparent with methods like that described in the article which make the refractive index uniform (i.e. exactly like snow that transitions from opaque to transparent when melting).
> I take out em dashes and replace them with commas or full stops
Yeah that's your problem. I like my em dashes. You don't need to be ashamed of your style just because it got mainstream. Doesn't matter whether it's AI or a fashion fad. It's like liking a music band before it gets popular on tiktok.
I agree it's a shame, but your comparison is confusing. Not wanting your writing to be mistaken for AI is not the same emotion as not wanting your writing to be interpreted as having "mainstream" or "fad" elements. They're completely different motivations.
It's the same emotion as if you're enjoying ska music subculture, but can't dress in its style of clothes, because people will think you're a skinhead.
You'll be judged for riding in your old Tesla car, because people may think you're a fan of modern Elon.
It may sound like a stretch, but I feel it's virtually the same thing — style.
If someone thinks my taste in music is lame because I like Dave Matthews Band, I'll wear that. But if someone thinks I don't care enough about the quality of my output to produce it myself, that's a more real and tangible hit to my reputation.
> You don't need to be ashamed of your style just because it got mainstream
It's not about shame, it's about your work being dismissed because it's an indicator that you didn't put any effort into it.
Although by the end, the article veers somewhat, and reads more like "I put a lot of effort into my writing, and just because I use AI also doesn't mean that I want to be lumped in with the people churning out AI slop". Which, okay, but that's not quite as noble a position as they seemed to start off with.
reply