Skip to main content

New top story on Hacker News: Show HN: Sonauto – a more controllable AI music creator

Show HN: Sonauto – a more controllable AI music creator
59 by zaptrem | 29 comments on Hacker News.
Hey HN, My cofounder and I trained an AI music generation model and after a month of testing we're launching 1.0 today. Ours is interesting because it's a latent diffusion model instead of a language model, which makes it more controllable: https://sonauto.ai/ Others do music generation by training a Vector Quantized Variational Autoencoder like Descript Audio Codec ( https://ift.tt/Iwk16oP ) to turn music into tokens, then training an LLM on those tokens. Instead, we ripped the tokenization part off and replaced it with a normal variational autoencoder bottleneck (along with some other important changes to enable insane compression ratios). This gave us a nice, normally distributed latent space on which to train a diffusion transformer (like Sora). Our diffusion model is also particularly interesting because it is the first audio diffusion model to generate coherent lyrics! We like diffusion models for music generation because they have some interesting properties that make controlling them easier (so you can make your own music instead of just taking what the machine gives you). For example, we have a rhythm control mode where you can upload your own percussion line or set a BPM. Very soon you'll also be able to generate proper variations of an uploaded or previously generated song (e.g., you could even sing into Voice Memos for a minute and upload that!). @Musicians of HN, try uploading your songs and using Rhythm Control/let us know what you think! Our goal is to enable more of you, not replace you. For example, we turned this drum line ( https://ift.tt/k0BAELe ) into this full song ( https://ift.tt/FrhqKok skip to 1:05 if impatient) or this other song I like better ( https://ift.tt/GrfPDFg - we accidentally compressed it with AAC instead of Opus which hurt quality, though) We also like diffusion models because while they're expensive to train, they're cheap to serve. We built our own efficient inference infrastructure instead of using those expensive inference as a service startups that are all the rage. That's why we're making generations on our site free and unlimited for as long as possible. We'd love to answer your questions. Let us know what you think of our first model! https://sonauto.ai/

Comments

Popular posts from this blog

New top story on Hacker News: Ask HN: Is the EULA on my new $30k RED cinema camera legal?

Ask HN: Is the EULA on my new $30k RED cinema camera legal? 70 by red_throwaway | 77 comments on Hacker News. TLDR: I bought a $30K professional cinema camera that doesn't work unless I sign away my rights to privacy and possibly the video content I make with it ( at least it seems ) Over the past few years my photography business has seen a surge in demand for ultra high quality video production work. In an effort to meet this demand, I picked up one of RED Digital Cinema's newest pro camera bodies, the RED V-RAPTOR. Considering this camera is used by professional filmmakers to create films destined for cinemas, it's not surprising that it came with a $30k price tag. After unboxing and assembling it, I power the camera on and the first thing I see is a wall of legal text on the embedded LCD. Turns out it's a "Software License Agreement" that I'm required to consent to using the on-camera menu buttons before any of the camera's functionality becomes...

New top story on Hacker News: Launch HN: Spoken (YC S21) – Better furniture shopping

Launch HN: Spoken (YC S21) – Better furniture shopping 14 by lgabraham | 3 comments on Hacker News. Hi HN, we’re Dane and Geoff, the founders of Spoken ( https://www.spoken.io/ ). We make it easy to find the lowest price for any furniture item across all big stores. Buying physical things on the internet is hard. You have to quality-check a product without touching it, double-check dimensions for where the thing will go, and evaluate a seller's credibility, often with little data. But buying furniture online is a special case of hard, because the market is deliberately deceptive. Furniture sellers actively prevent consumers from easily finding the same item at other stores, or under other names, because this allows them to charge more. The sellers get to name the products and they name them in confusing ways to facilitate price discriminaton. For example, this table at Wayfair ( https://ift.tt/RF071cQ... ) can also be found at Appliances Connection under a different name for rou...