search

mozilla / TTS

:robot: :speech_balloon: Deep learning for Text to Speech (Discussion forum: https://discourse.mozilla.org/c/tts)
Mozilla Public License 2.0
9.43k stars 1.26k forks source link

[HELP] Share your trained TTS models? #395

Closed erogol closed 3 years ago

erogol commented 4 years ago

Please consider sharing your pre-trained models in any language (If the licenses allow that). We can include them in our model catalog for public use by referencing your name (website, company etc.).

That would enable more people experiment together and coordinate, instead of individual efforts to achieve similar goals. That is also a good chance to make your work more visible.

thorstenMueller commented 4 years ago

Great idea. I'll provide a free to use german tts model of my own voice (tacotron v1 and v2). Training is currently running. After it's finished i'll publish download links on my github project page.

Samples of my original recording voice and "training-in-progress"-samples are here: https://soundcloud.com/thorsten-mueller-395984278/sets/thorsten-de-german-tts-model (currently just tacotron1 training on step 100k)

https://github.com/thorstenMueller/deep-learning-german-tts/ https://discourse.mozilla.org/t/contributing-my-german-voice-for-tts/

erogol commented 4 years ago

I added your great work into the list https://github.com/mozilla/TTS/wiki/Dataset

davidak commented 4 years ago

It is not a model, but a german dataset: https://www.kaggle.com/bryanpark/german-single-speaker-speech-dataset

stale[bot] commented 4 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discourse page for further help. https://discourse.mozilla.org/c/tts

thorstenMueller commented 4 years ago

I'd like to keep this issue open due i think it's important to collect available tts models centrally.

r-dh commented 4 years ago

Both my Dutch (nl-BE) dataset as well as several trained models: https://github.com/r-dh/dutch-vl-tts

thorstenMueller commented 4 years ago

Thanks for sharing your dataset and model. Your Taco2 + DDC has same robotic/metalic voice we encountered too.

German samples: https://community.mycroft.ai/t/free-german-tts-voice-for-mycroft-sneak-preview/9458?u=thorsten

ysujiang commented 4 years ago

Thanks for sharing your dataset and model. Your Taco2 + DDC has same robotic/metalic voice we encountered too.

German samples: https://community.mycroft.ai/t/free-german-tts-voice-for-mycroft-sneak-preview/9458?u=thorsten

can you share some generated voices?

thorstenMueller commented 4 years ago

Hello @ysujiang . The samples of generated voice are posted in the link above, but to summarize it up:

stale[bot] commented 3 years ago

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions. You might also look our discourse page for further help. https://discourse.mozilla.org/c/tts

Wingtail commented 3 years ago

Hi, I would like to share my Tacotron 2 model in the Korean language. It's currently training using forward attention. I've yet to do the vocoder, but I'm 200k steps in now and the sound is great.

Samples of generated voices in 200k step training: https://soundcloud.com/brian-kim-278776560/sets/korean-tacotron-2-forward-attention-200k-steps

erogol commented 3 years ago

You can either list in .models.json if you already keep it on GDrive or if you share the model I can handle the rest.

Hi, I would like to share my Tacotron 2 model in the Korean language. It's currently training using forward attention. I've yet to do the vocoder, but I'm 200k steps in now and the sound is great.

Samples of generated voices in 200k step training: https://soundcloud.com/brian-kim-278776560/sets/korean-tacotron-2-forward-attention-200k-steps

You can either list in .models.json if you already keep it on GDrive or if you share the model I can handle the rest.

erogol commented 3 years ago

Both my Dutch (nl-BE) dataset as well as several trained models: https://github.com/r-dh/dutch-vl-tts

would you like to add it to our .models.json as the first Dutch model?

erogol commented 3 years ago

Hi, I would like to share my Tacotron 2 model in the Korean language. It's currently training using forward attention. I've yet to do the vocoder, but I'm 200k steps in now and the sound is great.

Samples of generated voices in 200k step training: https://soundcloud.com/brian-kim-278776560/sets/korean-tacotron-2-forward-attention-200k-steps

Any updates on that? It'd be great to have Korean capabilities in TTS.

Wingtail commented 3 years ago

Hi, I would like to share my Tacotron 2 model in the Korean language. It's currently training using forward attention. I've yet to do the vocoder, but I'm 200k steps in now and the sound is great. Samples of generated voices in 200k step training: https://soundcloud.com/brian-kim-278776560/sets/korean-tacotron-2-forward-attention-200k-steps

Any updates on that? It'd be great to have Korean capabilities in TTS.

Hi, sorry for taking quite a long time. I've been prepping for midterms. I'll upload the trained model as soon as I'm done with it.

snakers4 commented 3 years ago

Hi,

Please see our TTS models here - https://github.com/snakers4/silero-models#text-to-speech (corresponding article https://habr.com/ru/post/549482/)

stefan-reich commented 3 years ago

Hello @ysujiang . The samples of generated voice are posted in the link above, but to summarize it up:

Is there a pretrained model to download? Would love to use this.

thorstenMueller commented 3 years ago

Sure @stefan-reich. If you've a python 3.6 - 3.9 environment ready it's really simple (see my github repo README): https://github.com/thorstenMueller/deep-learning-german-tts/#quick-steps-for-synthesizing-voice

stefan-reich commented 3 years ago

Sure @stefan-reich. If you've a python 3.6 - 3.9 environment ready it's really simple (see my github repo README): https://github.com/thorstenMueller/deep-learning-german-tts/#quick-steps-for-synthesizing-voice

  • pip install -U pip TTS
  • tts-server --model_name tts_models/de/thorsten/tacotron2-DCA
  • Open web-browser on http://localhost:5002

Ah, I'm dumb (uh, well, just sometimes ^^). Your model is literally in --list_models, I just hadn't checked out the international voices yet. Thanks. My only remaining desire is to have a female German voice because I may want to combine English and German words sometimes.

thorstenMueller commented 3 years ago

Sorry, but i "just" can provide a male voice :laughing:.

stefan-reich commented 3 years ago

Sorry, but i "just" can provide a male voice .

Is it a regeneration of your own voice? (Sorry, last question, I'll stop cluttering the thread.)

thorstenMueller commented 3 years ago

Yes, i am the speaker of the "Thorsten" dataset and i trained the models based on it. See: https://github.com/thorstenMueller/deep-learning-german-tts/

Vis5 commented 2 years ago

@Wingtail Hello, are you wiling to share your model if it's finalized? if not, can you point me in the direction of where you got your dataset?

souvikg544 commented 2 years ago

Please consider sharing your pre-trained models in any language (If the licenses allow that). We can include them in our model catalog for public use by referencing your name (website, company etc.).

That would enable more people experiment together and coordinate, instead of individual efforts to achieve similar goals. That is also a good chance to make your work more visible.

To create your own models first of all you need a dataset . Presenting you with my easy to use colab based tts dataset maker - https://github.com/souvikg544/TTS_Data_Maker

If you like it do give it a star. Any problem you can share on issues.