fishaudio

/

fish-speech-1.5

like 305

Follow

Fish Audio 302

Sounds good, do you mind sponsor Fish Audio $3M so we can continue our experiments and make it the "free" way you like?

Look at the foundational model and make it reasonable. Mozilla produces high quality software that way.

You are very strict, "I need money or I don't give it for commercial use", but are you not aware that by opening it, you may get actually the funds? You could even join your project with others in exchange for funds.

You are also building on a lot of truly free software, so you are gaining from free software, which can be used for commercial purposes.

Why would you restrict it?

You got not enough resources to go after those people who will use it commercially, in fact you cannot even know it! I can do anything with software, in fact I am on other continent, and I could build business with it, nobody would ask me anything about your rights.

What you cannot enforce, where you cannot get any gain, there is no reason to restrict it.

Free software does not mean free as in money, but free as in liberty.

You can monetize, sell, upgrade, your free software, once it becomes free.

You are already on the line of people where majority of the software is free software. You can only gain more by making it truly free.

Make clear practical services on your website, which will allow you to monetize and gain on it.

Research, there are so many ways to make money with free software.

You see, here is example, I was researching which TTS I may freely download and generate speech so that I may include it in sales and marketing. I am on Debian GNU/Linux and those built-in TTS programs are not sounding naturally. When I found the list, your software was among there, but I see -NC- tag and I let it be, I do not want to touch it, neither to research it.

And 1000 other people probably did the same, you lose contributors.

Then I went back to Mozilla TTS and installed it today, and I am now using various local models:

tts_models--de--thorsten--tacotron2-DDC       
tts_models--en--ljspeech--glow-tts            
tts_models--en--ljspeech--tacotron2-DDC       
tts_models--it--mai_female--glow-tts          
vocoder_models--de--thorsten--hifigan_v1      
vocoder_models--en--ljspeech--hifigan_v2      
vocoder_models--en--ljspeech--multiband-melgan

and I am satisfied with Thorsten and Glow TTS, and will research more before making final decision. Of course, I become part of project and help other people use it too! I have already written the script to provide stdin to the tts command as I did not find built-in solution:

#!/bin/bash

# Check if an output file name is provided
if [ -z "$1" ]; then
  echo "Usage: $0 <output_file>"
  exit 1
fi

# Prompt the user to enter text and read from standard input
echo "Enter the text you want to synthesize (press Ctrl+D when done):"
input_text=$(cat)

voice='tts_models/en/ljspeech/glow-tts'

# Run the TTS command with the input text and output file
/home/data1/protected/TTS/bin/tts --use_cuda USE_CUDA --model_name $voice --out_path "$1" --text "$input_text"

and I made Emacs Lisp package rcd-tts.el so that I can use it within Emacs and programmatically, part of it is here shown:

(defvar rcd-tts-command "/home/data1/protected/bin/rcd/rcd-tts.sh")
(defvar rcd-tts-output-directory "/home/data1/protected/tmp/tts")

;;; Basic functions for logging and files

(defun rcd-tts-log (text file type)
  "Log FILE as name and TEXT as description."
  (when (fboundp 'rcd-general-log)
    ;; only run if function `rcd-general-log' exists
    (rcd-general-log file text nil nil nil 1 type)))

(defun rcd-tts-file-name (text &optional extension)
  "Return TTS file name as UUID with optional EXTENSION.

Default EXTENSION is `webm'"
  (let ((b2sum (rcd-checksum-b2sum text))
    (extension (or extension "webm")))
    (format "%s.%s" b2sum extension)))

;;; Mozila TTS

(defun rcd-tts (text)
  "Return audio speech file for TEXT."
  (let* ((file (rcd-tts-file-name text))
     (audio (concat (file-name-as-directory rcd-tts-output-directory) file)))
    (cond ((file-exists-p audio)
       audio)
      (t (let ((result (rcd-command-output-from-input rcd-tts-command text audio)))
           (when (file-exists-p audio)
         (rcd-tts-log text audio 17)
         audio))))))

(defun rcd-tts-and-speak (text)
  "Speak TEXT."
  (let ((audio (rcd-tts text)))
    (when audio
      (rcd-mpv audio))))

and now, I am planning to make speech for various database functions, and I have 273 tables in the database, managing about everything.

What happens during that story is that I get more attached to Mozilla TTS, and then I also start promoting, and see use of it, and maybe donate, maybe not, but there is foundation that does receive money.

I don't say that model of foundation is easy for small companies, it is not.

But you could monetize!

Today I have made plugin for website, so the TTS output is generated and then audio file appears on the website page or it automatically starts playing. I have 1500+ pages, but someone else has so much more.

You could be making website pages talking. Isn't that good thing? Small subscription and you could generate for each page:

• summary description
• shortened version of the page, talking nicely

and visitors could turn on talk while browsing. You could help with the automated integration.

Imagine letting users insert some HTML code, so that the HTML code detects the page, and automatically grabs your audio to show it on the page. Users would need to change template only one time, and all website pages become talkative.

I am sure there are many good ways to monetize, but it requires sales and marketing skills, not just scientific and computing skills.

You could hire someone to start working on such projects.

This idea sounds fantastic! Would you be interested in investing 2.5M CNY as seed funding for what could be one of the most groundbreaking projects in history?

I appreciate your enthusiasm, but I'm not in a position to invest such a large amount. I think the focus should be on making the model commercially viable first.

Oh, you wrote a nice little essay there - how cute. But you know what? Talk is cheap.

Let me drop some reality on you about commercialization:

• The SECOND we remove that "no commercial license" clause, vultures will swoop in to profit from it. Don't believe me? Just look at what happened to SovitsVC and Bert-VITS2. Their creators? Not a single penny.
• Remember XTTS? That "famous" TTS project? They're dead in the water now because they couldn't make a dime. Their creators literally had to abandon ship.

You're comparing us to "small TTS models"? That's adorable. We've dumped nearly a million dollars into Fish Speech. We're keeping it open source so people can actually use it - you know, our actual mission. And since you brought up Mozilla - while they do great work, let's be real: over 90% of their revenue comes from search engine deals. That's just the reality of sustaining open source projects.

Look, I get it. Maybe this doesn't fit some business models. And yes, there are genuine open source idealists out there doing great work - the community needs them and we respect that. But what we don't need are armchair critics who just run their mouths without ever contributing code.

But congratulations! Your stunning display of ignorance has so thoroughly disgusted our core team that we're now pumping the brakes on our open source plans.

Wait you are letting one person with a request to make the model actually free and open hinder the entire project? what

Also the comments from the fish team seem very condescending and rude. Just because you don't agree with that persons statement doesnt mean you should talk down to them. And asking them repeatedly for money is quite unprofessional.

Oh look, another issue like this. You know what's fascinating? This is actually the first time we've responded negatively - and only because the original post was dripping with that special kind of condescension that comes from people who love preaching from their moral high ground while being completely detached from reality. And here you are, following the same playbook.

You're clutching your pearls about us mentioning costs? That's rich. Maybe spend a minute learning about what it actually takes to develop ML models before jumping on your soapbox. And yes, I mentioned funding with a touch of sarcasm - but aren't you here taking potshots at us too?

You know what? This whole exchange has done nothing but cement our position. We just had a chat with all our top contributors - you know, the ones who actually have internal access and make sure updates roll out smoothly - and guess what? They unanimously agree we should hit pause on open-sourcing.

So let me be crystal clear: this isn't a suggestion up for debate. It's a notification.

PS: The irony of lecturing us about professionalism while misreading the entire situation isn't lost on us.

Did you just paste my comments into chatgpt and ask it for a snarky response? Ill be real with you, that reeks of large language model.

I have no investment in this project. You could wipe it off the earth for all I care. But be nicer to people. Thats all I want to say.

I appreciate your concern about communication style. But this isn't an AI response - it's genuine frustration from a team that's been quietly managing both investor pressure and community expectations while trying to make TTS / AI Companion accessible to everyone. Opening up Fish Speech hasn't been easy, but we've never brought these challenges to the community until now.

I saw your GitHub history - you've made valuable contributions to open source over the years, so you should understand the complexities involved. That's why it's particularly disappointing to see you focusing on tone rather than substance, especially when you haven't contributed to this specific project.

We're working towards making AI companions accessible to everyone. That means making difficult decisions about resources, funding, and sustainability. These aren't just theoretical concerns - they directly impact our ability to deliver on our mission.

You don't have to have any investment in this project - that's fine. But we do, and we're committed to making the right decisions for its future, even when those decisions aren't popular.

I understand where you are coming from, but the ChatGPT responses are painfully obvious and you aren't fooling anyone. The only reason I chimed in was because your previous comments were a bit harsh to the original poster.

Its your project, of course you dont have to listen to OP, but halting open source development because of that is sort of abnormal. It caught me off guard along with the responses you are giving.

We will also open source later, but the nc protocol will not be modified in a short time. We have great respect for those who really strive for the completely free open source cause, but now survival and sufficient funds are the main goals. We have invested a lot of money in it. We don't want thieves to steal our achievements directly. At the same time, we also hope to have more funds to support us. You know, now related types of research and development need a lot of money to operate normally. Judging from other non-nc TTS projects in the past, most of them can't survive normally and stably. This road is far from the end, and we must have enough funds for research and development to better support the cause of open source.

Oh, you wrote a nice little essay there - how cute. But you know what? Talk is cheap.

It was meant to give you enlightenment, an example, how user jartine got a grant from Mozilla.

No, I have nothing against being commercial! That is the point, you must be commercial and you are building good website that was basically same thing I was thinking how to do it. I know how much effort is required to market, promote, especially in this now emerging and very competitive market.

I am of opinion that we are all commercial, we must be, otherwise how to make it in life? Unless one is a humble farmer with the land, one need to provide goods and services to exchange with others, and commerce is alright, it is exchange, we all need it.

Let me drop some reality on you about commercialization:

• The SECOND we remove that "no commercial license" clause, vultures will swoop in to profit from it. Don't believe me? Just look at what happened to SovitsVC and Bert-VITS2. Their creators? Not a single penny.

I am not acquainted with them. But you must understand that a lot of companies will not care of your licensing issues, as long as they can get the code. They will take your software and use it commercially if they want, and you will not know it. Don't think that bad guys play by the rule, they don't. I hope you understand me. That Creative Commons license version forbidding commercial use doesn't prevent such cases. I don't find it good for software.

Maybe it is good for images, put my images on your website, but please don't sell them, and I don't want you to use my images to sell your products. I can get that case. Let's say learning courses online, learn as much as you want, but please, don't sell courses as I want education to be truly accessible and not monopolized. Let's say music, play it as much as you want, but please, don't sell, it would hurt my feelings.

Do you see there some types where I can understand authors using CC-NC license.

In the case of company wishing to make money, you need to get some advantage, something useful and good for customers, so that your product is recognized and that you are above the competition, at least that you are among few who can do the magic.

Example: I have just downloaded Mistral Nemo llamafile model and it runs on my low end CPU giving me quality output. I am happy with it. And I can see that Mistral organization trains models for money, so they got the platform, and they got their hardware or outsourced providers, organized the business and are distributing free models which in turn bring them customers for the tasks that common people can't do on their hardware, such as training the model on specific information. And for sure I will contact them soon, as I have bunch of specific knowledge books on which I would like Mistral model get trained. Distribution of free software, brings them customers for their high-end products. That is strategy of Mistral.

Personally I have made business management software and Dynamic Knowledge Repository, and my clients are eager to have it as soon as they see it, so I am making money on different income line such as manufacturing of machines or consulting them on business startups, but in the same time giving them free software to use in their own management by the developed and organized managed style. I can also sell software, provide support, and still have it free software.

I think if your only reason to make it CC-NC is to prevent others earning money is that it is waste of time. Because no matter how skilled you are in international law, it is very difficult preventing others, especially in different countries on different continents.

We talk about the TTS, after I looked at your page here, I installed Mozilla, then I installed piper-tts, and that one is even faster. I didn't look at your software, as in general I will not download non-free software. All the GNU/Linux users are mostly running free software on their systems, they have different mind set. All the Hugging Face is built on lot of free software, all of the context is about free software. It is different community.

• Remember XTTS? That "famous" TTS project? They're dead in the water now because they couldn't make a dime. Their creators literally had to abandon ship.

Many businesses fail.

If you think to make business with fish-speech and that the way is to forbid commercial usage, feel free. But I am giving you my experience, and some pointers. You could look at those models of making money by having some advantage, and you are already there, as I have peeked at your website, I am sure you are on the right path.

Locally providing speech services to companies, generating TTS for companies under contract, that is something I would be doing with my imaginary speech software, if I would have it. You could let software of other companies speak, that is in fact reason why I have downloaded it, as when I browse the Dynamic Knowledge Repository, I just change the function and get better output than Festival or Espeak. Though software is used in group, in the team, where people watch and participate, do financial reports. In combination with LLM generation of the report, I can let reports be spoken, it adds to the experience. Same similar services you could provide to local companies.

Believe me, there is large number of people who can't read. They say they can, but in reality they don't. They watch TikTok. Speech will be more and more necessary in our ever stupid-becoming society.

You're comparing us to "small TTS models"?

Not my intention, I said I know nothing about it. It is just experience that I didn't download it due to licensing.

That's adorable. We've dumped nearly a million dollars into Fish Speech. We're keeping it open source so people can actually use it - you know, our actual mission. And since you brought up Mozilla - while they do great work, let's be real: over 90% of their revenue comes from search engine deals. That's just the reality of sustaining open source projects.

Sure, that is another model of making money, maybe they are getting it from Pocket, or other sources, or search engines. Can't know. But that is to be taken as good example, and software you publish as free software, let's say here on Hugging Face, shall not be your main method of making money. Make it with other different advantages, don't rely on making money on free downloads, make those downloads a bait for your other money making model.

Look, I get it. Maybe this doesn't fit some business models. And yes, there are genuine open source idealists out there doing great work - the community needs them and we respect that. But what we don't need are armchair critics who just run their mouths without ever contributing code.

You see I am free software user, so I help some people and make packages in different types of software.

As you said, you like people who contribute code, so there we are! That means you understand the point of free software.

But it is kind of contradictory to contribute code which you will use for proprietary reasons, whereby contributor should disregard all the envy you got against possible perpetrators, and he shall contribute to software where you will be making money, but contributor not...

But congratulations! Your stunning display of ignorance has so thoroughly disgusted our core team that we're now pumping the brakes on our open source plans.

I am not user of "open source" as the term is diluted with a lot of proprietary software pretending to be free software. I am user of free software only, never used "open source", that is branded term for reasons to lick the asses of large corporations.

No, I am not ignorant to your business, as I am business consultant. I just think that your main income line should be strengthened, and should not depend on any kind of publishing of your software, for reasons that you can't really prevent people perpetrating it if it is already published.

Train your own model plz