License terms: what about using this in a commercial product?

juangea commented 2 months ago

Self Checks

[X] I have searched for existing issues search for existing issues, including closed ones.
[X] I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
[X] [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:）
[X] Please do not modify this template :) and fill in all the required fields.

1. Is this request related to a challenge you're experiencing? Tell me about your story.

It's related to the license terms, and how to use the app commercially

2. Additional context or comments

My question is about using this commercially.

I don't mind about Attribution, the Share Alike, I may not mind either, since I don't want to modify the core and I don't mind being clear about what we are using, and it would be included in the environment of our app.

However the NC part makes this unusable for us, we are a small company that is trying to do some useful local-based tools, money speaking we don't have much, so we try to use as much open source under MIT or Apache2 we can, right now for TTS we use Bark, but the quality could be better with FishSpeech, I wonder if there is a way to get permission to use FishSpeech for our commercial apps, we have no app released yet.

3. Can you help us with this feature?

[ ] I am interested in contributing to this feature.

irux commented 2 months ago

I would love to have this open source but probably it is NC because the data set is trained on. I think they will not change the license to be honest and if you need the comercial version, probably you would need to use the fish.audio website for that.

juangea commented 2 months ago

Thanks for the answer.

That's a pity since our target is to focus on local/private ai developments, avoiding any kind of third party api usage.

WE would be using XTTSv2 but since the licenses cannot be acquired anymore it's impossible, and other models like MeloTTS, that support several languages (which we need) have a mid to low emotional quality, bark is the best one so far as long as we use it with RVC afterwards, but it's not too stable and the emotion and tone are not really under control, it does what it wants to be honest.

So this model seemed pretty interesting with very good result in several languages, and with good emotional tone and prosody, but the license makes it unusable, what a pity.

I hope the authors can answer about this and clarify the situation and/or possible changes, but thanks for your answer :)

irux commented 2 months ago

Well there is always many others out there that are great and open source too. Try to find useful things on papaperswithcode. There is always the option to train your own model and try to find the paper this is base on and replicate it, maybe perhaps a clean room implementation? (not a lawyer and not legal advice btw). Obviously expensive and more difficult, but obviously it is understandable that the team want to have comercial gains with this. Other option is to look at the version before 1.0 where at least the code was BSD.

juangea commented 2 months ago

@irux thanks for your answer, as I said we are very small, we have to rely on open source projects or projects we can afford, so training our own model is out of question I'm afraid, as for other projects, the only multilingual projects with some quality are Coqui with XTTS that I already mentioned, and Bark, others like OpenVoice sound more or less good on english but are absolutely terrible in spanish for example, that's why I'm actively looking for more multilingual projects.

I would like to ping @leng-yue since seems to be part of the team.

Actual real clarification on this is very welcome if possible.

Thanks!

leng-yue commented 2 months ago

We're considering open-source options like AGPL or a capped license for smaller users. However, as a startup investing heavily in model development, we're unlikely to offer a fully open commercial license.

juangea commented 2 months ago

@leng-yue I totally get it, AGPL could perfectly work, that would allow small business to use it commercially even for cloud services as long as they (we) provide your source code and we do not deeply integrate it in our software, but use it inside the server as an I/O by file solution, and will prevent big companies from integrating it without providing you anything.

Also you can opt for a fully commercial license, as an option for small business and mandatory for business over 1M revenue per year, this way once again small business can use the software and if they grow they should pay you a small percentage of the revenue, like 2% or 3% per quarter from the first 100k$ onwards.

And you can limit that license to companies from 1M to 5M, and from 5M onwards they should acquire a license with specific terms, there you define specific terms company by company.

I think this could ensure the viability as a business now and in the future, enlarge your user base and ease the integration of your software in startups and small companies that can be big companies in the future, and at the same time you are providing something to the OSS community.

This situation could be a win win for everyone IMHO.

Please if you decide to enable an AGPL license or something similar, make it clear and loud, I would like to use Fish Speech but as it is right now we can't do it.

juangea commented 2 months ago

@leng-yue when you make your mind around this please inform ASAP, I really would like to use this, it's better than bark and I love the way it picks the tone and amotion from the source audio, something that it's very hard with Bark, but with the current license it's impossible, with AGPL or similar it would be totally possible, I hope you give us a solution for this soon :)

Thanks!

leng-yue commented 2 months ago

We will update in this issue when we make some decision :)

platform-kit commented 1 month ago

+1 on a commercial use friendly license.

lukaLLM commented 3 weeks ago

+1 also

fishaudio / fish-speech