Silero Models License Infringement

[snakers4]

Hi,

My name is Alexander, I am writing to you on behalf of Silero, a company maintaining our project silero-models.
We noticed that our models are hosted here - https://huggingface.co/spaces/pytorch/silero_tts or here https://huggingface.co/spaces?search=silero.
We did not explicitly grant Hugging Face, Inc. any sort of permission to rehost, relicense and profit from our work and models.
Moreover it openly disregards our CC BY-NC-SA license.
Please immediately remove any of our models from your website and / or any of your resources.

Best,
Alexander

[julien-c]

Hi @snakers4,

Thanks again for your report, we've taken a look promptly.

We've added an explicit mention about the CC-BY-NC-SA license in the Gradio interface, see this commit

For context, we do not host the model, but simply load it from PyTorch Hub: https://huggingface.co/spaces/pytorch/silero_tts/blob/main/app.py (PyTorch Hub source: https://pytorch.org/hub/snakers4_silero-models_tts/)

Would this explicit mention work for you? Otherwise, we're happy to delete the Spaces.

cc @annatrdj @osanseviero

[snakers4]

Hi @julien-c ,

Otherwise, we're happy to delete the Spaces.

Please do.
Many thanks.

Best,
Alex

[julien-c]

Hi @snakers4, we deleted the Spaces

Ping us if we can help in any way in the future!

[snakers4]

Hi,

@julien-c

We just noticed that our models are re-hosted by you again:

Please delete them as well.

Also there is a lot of parasite traffic constantly downloading these models, which may be related to this.

[julien-c]

Hi, i would suggest you contact model uploaders in the Community tab of each model.

That being said, if i understand correctly CC BY-NC-SA explicitly allows redistribution of materials no? (under the same license, of course)

[snakers4]

Hi,

That being said, if i understand correctly CC BY-NC-SA explicitly allows redistribution of materials no? (under the same license, of course)

I see no CC BY-NC-SA license text here at all:

• https://huggingface.co/imperialwool/silero-model-v3-ru/tree/main
• https://huggingface.co/aguilara42/silero-vad-wrapper/tree/main
• https://huggingface.co/neongeckocom/silero-lang95/tree/main

Also I see no attribution whatsoever.

Hi, i would suggest you contact model uploaders in the Community tab of each model.

I would suggest instead that you would immediately stop infringing on our license, profiting from our non-commercial releases and re-hosting our work without any proper attribution.

Please delete these re-uploads immediately.

[hugofloresgarcia]

Hi!

I’m writing on behalf of the Audacitorch team (https://interactiveaudiolab.github.io/project/audacity.html), where we work to get open source deep models in music production software for free, aiming to bridge the music production and audio deep learning communities. As a team that works towards integrating open source deep models into music production software, we strive to contribute to the community and support projects like yours. We acknowledge that our oversight must have caused inconvenience, and we take full responsibility for our mistake.

I’d like to apologize for not including the correct license and attribution for our repo, we did not intend to infringe on the licensing of your work. We understand the importance of this issue and would like to rectify the situation as soon as possible.

The reason we had to reupload the model to it’s own repo in the first place is because we had to wrap it with our own code to make it compatible with audacity (see https://github.com/audacitorch/audacitorch), and have it available on huggingface so end users can pull the model into their own Audacity environments.

We’ve corrected this immediately and uploaded a copy of the CC BY-NC-SA license to our repo (see here: https://huggingface.co/aguilara42/silero-vad-wrapper/blob/main/README.md), and included all other attributions described in the license, as well as included attributions in the user interface. Let us know if there’s anything else we’re missing, and we would be happy to correct.

If however, you would still like for us to take the models down and delete the repos, we would be more than happy to do so.

Thanks!

[snakers4]

The reason we had to reupload the model to it’s own repo in the first place is because we had to wrap it with our own code to make it compatible with audacity

I am not sure why downloading the model file directly from the official author repo does not work.

This attitude shows the apparent blatant disregard for the authors of open models and disincentivizes any further work, where non-commercial third party releases are used as marketing tool by for-profit venture-backed institutions, the fig leaf being that "look - community did it, not us".

This also bears some resemblance on how the image generation apps treat the authors of the digital art.

In both cases, it is clearly a mockery of the open source principles / decency and will have long-lasting consequences for the industry.

If however, you would still like for us to take the models down and delete the repos

Please do.

[julien-c]

@snakers4 I don't understand why you chose this license then https://creativecommons.org/licenses/by-nc-sa/2.0/ – seems a very different license from what you are trying to achieve. You should disallow re-distribution, no?

[snakers4]

@julien-c
I am not against non-commercial redistribution.

Here our model is used as a marketing tool by a for profit company, which designed its platform to benefit the original authors as little as possible.

[hugofloresgarcia]

I am not sure why downloading the model file directly from the official author repo does not work.

It's not the same model file. We took the model and wrapped new pytorch layers around it to be compatible with Audacity. I would be more than happy to talk more into details about this!

This attitude shows the apparent blatant disregard for the authors of open models and disincentivizes any further work, where non-commercial third party releases are used as marketing tool by for-profit venture-backed institutions, the fig leaf being that "look - community did it, not us".

We're (audacitorch) not a for-profit organization -- we're a university-affiliated research group that aims to bridge the gap between deep model developers and users of audio production software by creating a generic end-user interface for a variety of Audio deep models -- our work is meant to be free and open source, and is backed by the non-profit universities where some of us are students. If you'd like to read more about our project, feel free to take a look at our paper: https://arxiv.org/pdf/2110.13323.pdf.

Again, just wanted to apologize for our mistake, and make our project's goal // stance clear. We've taken the model down.
Not including the correct licenses was a beginner mistake on my personal end. I sincerely apologize.