fishaudio

/

openaudio-s1-mini

like 336

Follow

Fish Audio 677

& also all of your credits are open source so it isn't a base model thing, I don't think:

Credits

VITS2 (daniilrobnikov)
Bert-VITS2
GPT VITS
MQTTS
GPT Fast
GPT-SoVITS
Qwen3

All either MIT or BSD

OpenAudio with NOT open license. ha ha

First, we do not have the duty to open source any weights and maintain the codebase.
Second, the codebase's apache license is for those who have the ability to train the model from scratch to use for commercial use.
Last, we spent millions of dollars to train the model while someone just want to steal our efforts with ease😅.
If you can train the model with the current codebase, then you can use it for commercial use.
For those who only take pleasure in the fruits of their predecessors, I advise you to shut up.

... A little hostile, no?

I really wasn't trying to offend you, I was just looking for clarification on licensing, so I'm sorry if that somehow seemed like I was making assumptions about how this research was done, who did it, why, and what people are using it for.

That being said... nobody said you have a duty to open source anything, but you CERTAINLY don't have a duty to call your project OPENaudio. Go look at every other popular project right now that has open in it's name, if you want, huggingface is full of them, mostly mit. But for a researcher from the team to pop off and attack the first person that even so much as INQUIRES about it after YOU are the one who named it that? Surely it isn't ridiculous to consider that it warrants a little criticism?

In short, be whatever research entity you want, but every sentence you just typed is literally the exact opposite of open, so I'm just offering feedback that maybe it's a confusion naming methodology if it isn't what you believe in or how you want your lab or whatever to be treated? Can hardly really blame someone for trying to "steal" an "open" model? You have to see how this is all a bit upside-down, right?

Dunno about everyone else, but I'm a disability aid researcher who was pretty stoked about this for folks with disabilities to be able to expressively communicate with augmented and alternative speech. I can't use it though, because of its licensing. Is that what you meant by stealing? What is the intended use of this, you just want it to be your API or die or? Just so we all know how to respectfully use this without stepping on toes.

how can you not use it ? you can you just have to attribute for it - and not sell a product
albeit potato tooked it a bit too personal - but there is nothing inherent that hinders you from using the model as is - apart from that you cant sell it as product

It does make sense that you would think that upon first impression, but unfortunately a lot of applications require a pretty sad amount of resources to adapt to. When trying to create solutions accessible enough for disabled folks, and even more difficult, into their possession, is a brutally costly job with a tough financial proposition. You don't really do it for the money. A lot of the time unless you're making a custom implemented solution (extremely costly one off purchased by the user) or it's something that a commercial entity might back, it doesn't get made, period. If it isn't commercially viable, the tens of thousands of dollars it would to make amazing technology like this accessible enough for real end user downstream applications would have to come from the dev. In other words, for it to actually be useful to anyone "in the real world" or get in any regular person's hands, It can't be NC. My mother doesn't know what an API is, and the beauty of this model is its efficiency anyway, it's begging to be run locally!

Things like art installations, experiments... trying to think of other uses off the top of my head isn't going well... that's fine for NC. Maybe if you have the time and skill to write a grant? I'm talking 10s of thousands here, not millions. We're small fish. I'm not making any statements about who the team is or what situation it would put them in to open source. What I AM saying is that, I kid you not, I am in a situation where I cannot use it professionally to try and make the world a bit better, and I would if it was truly open. The magic, a lot of the real differences that end users get to have in their lives, the reason why a lot of us do why we do, It just isn't possible without free and open source software, and free as in libre, not gratis. It really is all up to people who make more money than I do in the end, but there is lots of literature (lol!) on why the four freedoms or principals of open source software are ALL needed for software to be open.

Again, all that aside, why would anyone spending money on making a serious system, serious enough to do something truly amazing with this incredible technology, think twice about this option? There's so many amazing options that parody ElevenLabs at this point that are fully open, truly permissive. If I was a boss, I'd say "nah, don't be so picky just get something open source" too, even if it was marginally worse. That stinks, because it's beautiful work, right? So like I was saying earlier, the only option it kinda boils down to for the team who made it at that point is to become yet another API provider. And again, if that's what you wanna do, get yours and you deserve it, you baked the bread and did the work yourself, but it ain't open.

i think if you run a biz you try to protect your assets too - while im not affiliated with fish - i do quite a bit in oss / and research in the audio domain + pour my private capital in it without any plan to recoop it .. most of us do it because we deem it important that it gets done - some want to make it a biz as they have to answer to investors - what ever it may be - you have options

i agree that openaudio as rebranding maybe a bit "cheeky" but then again so is "open ai"
if you need fully open .. give chatterbox a go - its close / csm is good too / orpheus isnt bad either ..

Unfortunately most people don't have private capital.

I've also said multiple times:

I know there are many open source options
I am supportive of companies making money off of their work and their freedoms to do so and to protect their IP.

And listen, you are DEFINITELY not going to hear my defending OpenAI, that is for SURE.