AI-generating music app Riffusion turns viral success into $4M in funding | Hacker News Reader

AI-generating music app Riffusion turns viral success into $4M in funding

techcrunch.com

60 pointsbemmu51 comments

Related: Show HN: Riffusion with Lyrics (https://news.ycombinator.com/item?id=37914425)

[dupe]

More discussion on show HN post over here: https://news.ycombinator.com/item?id=37914425

I listened to like a dozen of the examples on riffusion and they all sound like junk. It's kind of like the hands being messed up on generated images, except the "messed up hands" stream through every part of each track.

Come on... we are very early in this new world. It's always funny the lack of imagination people have with this stuff. Have you not been observing the rapid progress of LLMs? This app was not possible a few years ago. What will these products look like a few years from now? What commercials on the Super Bowl will be using AI generated music in the background? Will the NFL Network be using these AI sounds during their transitions?

I mean, yeah most of these are lame. I found one or two that actually sounded good. But they are short clips. It's probably going to take awhile for AI to generate an entire 4-minute track cohesively.

I understand your sentiment but I have to agree with the parent. Sure, probably in a couple of years it will get better. But for now, it simply isn't good. Even generative music from the pre(current)-AI era sounded much better, even though it was quite boring.

Really, I am interested, I'm looking forward to new developments, but at the same time I have the right to call a spade a spade, to be euphemistic.

...so? The parent didn't say music generation AI will never be good. They were just commenting on one single thing, which is Riffusion.

I have no experience in this area, but I feel like the ideas are the hard part in music. These are rough drafts and some of them have promise if they can be refined in an iterative fashion with the user at the helm.

You can generate very interesting music simply by working with MIDI as opposed to sampled audio (slashing complexity by orders of magnitude!) and starting from a good model architecture. Daniel D. Johnson's (formerly known as Hexahedria, hired by Google Brain) model biaxial-rnn-music-composition is from 2015, requires very few resources for training or inference, and still delivers compelling, SOTA-or-close results wrt. improvising ("noodling") classical piano. Github https://github.com/danieldjohnson/biaxial-rnn-music-composit... , you may also want to check out user kpister's recent port to Python 3.x and aesara: https://github.com/kpister/biaxial-rnn-music-composition (Hat tip: https://news.ycombinator.com/item?id=30328593 )

"There’s no clear monetization strategy — yet. For now, Forsgren and Martiros say that they’re focusing on growing Riffusion’s team and developing complementary new generative AI products."

Do burn that silly VC money like there is no tomorrow!

The world would be a much better place if everyone who took VC money just then gave it to something worthwhile.

Why couldn’t they license the content you generate using their service such that if your revenue for that specific song goes above $x, they reserve the right to a y% cut?

To be fair, my own musical projects and every one of the 50 or so bands I've made money playing with has had the same business plan.

I am sure that this effort will be similarly successful.

I am not following how playing in a band for money has the same business plan as this app

I was being a bit tongue in cheek. However this was relatable, especially to the plans of the original musicians:

"There’s no clear monetization strategy — yet. [...] they’re focusing on growing [the band] and developing complementary new [songs|social media].""

It would be funny if the credits were branded as "drink tickets" when you generated something

https://www.riffusion.com/riffs/10308aee-b9a1-4f08-8236-fa86...

Buy that lambo, you made it :)

https://www.riffusion.com/

Reality is getting crazier day by day.

how i hate exotic apps that don't explain what they do but just present me some buttons to goof around

I would register if I didn't have to enter my phone number or login via Google.

I'm more impressed that there was zero errors in the console on loading. impressive.

This kind of stuff is poised to absolutely takeover for video background music.

Dealing with copyright, royalties, licenses, etc. is such a pain in the ass for a lot of small creators, and the ability to just create something totally new is a game changer.

Yep. Just navigating trying to find music you can use for an Instagram reel advertisement (or the like) is incredibly frustrating or expensive.

It's interesting that you implied if you use AI you are free from "dealing with copyright, royalties, licenses".

That's the only real use of AI. IP laundering.

The US Copyright Office has repeatedly refused to register copyright for substantially AI-generated works. There is a separate question whether the creation of the model could infringe on rights in the source material, but so far, usage of a model to create a work appears to lead to a different conclusion. Of course, this is a new area and subject to change.

https://arstechnica.com/information-technology/2023/09/us-re...

Interestingly, a service that made that use-case into a turn-key and repeatable API call would probably be quite profitable. The target market would be web marketers and social media managers who have to pump out content.

I'm imagining something like: upload a video without sound and the service would generate the exact amount of music necessary. Bonus points if you can get the music to loop seamlessly. Perhaps upload an image or series of images + text and the video would be ffmpeg'd together at a specified duration along with the generated music. Add some integration to manage the uploading to Instagram/TikTok/Facebook/etc.

It reminds me of a business I saw launched here that was an API to remove backgrounds from images. I didn't think such a service would be profitable but apparently there are enough people generating product photos for online stores that will throw money at such a service.

Take it one step further and create an audio track tailored each viewer. Simply read the user's tastes and marketing profile and generate something that appeals to them. If they have a preference for rock, add some guitar. Country would get a bit of twang, jazz could add some brass. It's likely no two viewers would get the same audio.

hahah where is the background removal service. i'm legit looking for something like that.

It feels like the time when the West was flooded with cheap plastic from China - because it was cheap.

Listened to three samples. Glad that my Roland synth has nothing to be afraid of.

I think this is a great usage of the technology that has the potential to produce very interesting and unique mixes. AI image generation has this ability to make images that at a glance create in your mind the prompt but then fall apart at the seems when you focus on any particular element. But for music "sounding like music" is the highest ideal it can achieve. You spend only a second on any particular part so your mind just fills in the gaps and ends up not being uncanny since "playing notes outside the frame" is already widely used in modern composition.

The lyrics are weird and it is gonna fall down on musical styles where structure is both more apparent and important but whatever, just making the backing track itself is a huge accomplishment.

Agreed - I was thinking if the lyrics were in Spanish, it would sound like legit music to me. The lyrics are kind of silly - like someone making up a song in improv comedy.

I've been interested in something like this - a sort of ephemeral perpetual MIDI LLM jam session. These clips seem short, I wonder if there's a longer form audio if you log in.

The investor hype in AI reminds me of the bitcoin hype around 2017. There is some real, promising tech, some real show- and usecases, but still ... Something's already off before it even started.

The fundamental issue isn't that some technologies are "off", lots of high-level research into things like mathematics and tech has absolutely no direct practical application. Lots of things are done with the mindset of "planting trees even though we ourselves will never be able to enjoy the shade".

The actual problem is the financial madness that's overtaken tech. Bitcoins is a great concept, having a decentralized system not owned by visa for doing payments across the planet would be a huge boon to a lot of people. But it drowned in how much money could be made by treating it as a financial investment. The same thing is happening with AI, and it's not the fault of AI itself as a tech. And frankly, all the comments about how "AI isn't all that, it won't be able to quickly upend every industry on the planet and make us billions" is a disappointing mindset to see. The hacker mentality has been replaced by middle-managers and financial experts.

If your complaint is simply the money that tech is swimming in, I have some bad news for you about the state of the Internet since 1998. First there was the web boom (and bust) (which was supposed to change the world). (Arguably it did, but eg record labels are still around.) Then there was the the Web 2.0 boom. Then there was crypto, now there's AI. Outside of tech, the world still looks largely the same. American "capitalism" is still filled with inequality, with lots of misery on all levels, and we don't have any more free time than we did before (unless your startup lottery ticket pays out).

Sorry for letting my cynicism show, it's pre-coffee, but you'll forgive me for thinking this time it'll be different. It might be, but we've heard that promise before and should be a) just as skeptical but also b) just as enamoured. If you missed out on the crypto boom (and subsequent bust), you've got some catching up to do in this cycles boom (and bust).

First of all, it's not at all similar to crypto. And secondly you are free to not invest in AI. Why is that every time AI company is discussed in HN, this topic comes up. This company has nothing to do with what created this investment trend.

Let's discuss the company and product and not low effort and political AI comment in general.

It is similar to crypto in that it’s relentlessly hyped by VCs and the product quality does not match the hype or funding.

The article is literally about the fundraising so the comments about the fundraising are more on topic than your pointless dismissal of them.

Counter argument: Every crypto coin is, more or less, the same. Sure there's some underlying difference in how they work (proof of stake vs proof of work) but they don't do anything wildly different.

AI companies, on the other hand, ARE fairly different in the products they're offering. So while it can make sense to talk about the crypto world as a whole when on a thread about an individual coin, that makes less sense when discussing individual AI companies IMO.

I really don't know of anyone using crypto other than for investing and get rich quick scheme. Crypto was overvalued as investors mostly have vested interest in coins. And funding startups is the cheapest way to inflate the coin value. And AI is solving real problems for me.

Let’s say I am the opposite and am very impressed by AI.

How exactly would I even invest in AI right now as a layperson? It seems like any investments would be either just straight NVDA or Microsoft and hope they eat the world. Since I’m not an accredited investor.

VCs have really destroyed people in tech's ability to think about and evaluate technology.

At this point I can't count the number of times I've heard someone say the equivalentof "VCs hyped toothpaste and orange flavored gravy. And now they're hyping sliced bread - they're the same thing!"

Stop deciding if something is good or not based on VC hype. Evaluate if the actual tech is good or not - if it does promise or not. Stop following VCs whims.

There sure won't be any AI metal songs, they even filter out the word "angry"..

I think the biggest issue with AI generated art is that it allows corporate folks (marketers, etc) the ability to create exactly what they want without any involvement/pushback from artists. Often the agenda of artists is do create something nice for everyone, whereas corporate folks only want to extract as much wealth as possible. Without artists in the loop, we'll only get more and more of the corporate agenda.

Things are complicated though. So remains to be seen how this evolves, :shrug:

I think it will be a problem for the corporate types too. Getting exactly what you ask for can be dangerous.

I’m a screenwriter and on a given project I’ll interface with a producer (a semi-corporate type) and a studio exec (a very corporate middle manager type). A big part of my job that AI can’t do any time soon is lying to them and ignoring them. I’ll get instructions (often contradictory instructions) from both of them, and then lie to them about how good their idea is. They feel listened to, but I’ll mostly ignore both of them so that the end product will be something actually good. At the end of the process they happily take credit for the good product and even believe it was their ideas that got it there.

No model?

I wish them the worst. I can see exactly no scenario in which them making money from this ends up being good for anyone besides them.

They publicly released a model a while ago, but I'm not sure if they use a different one now.

I meant "business model" here.

Reply on news.ycombinator.com