Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Repo info
Activity
    Edresson ๐Ÿธ
    @edresson:matrix.org
    [m]
    1 reply
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    You mean flacs?
    1 reply
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    just checking because there are different versions of the data floating around ๐Ÿ˜…
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Edresson ๐Ÿธ: is there a length cutoff in the training code?
    1 reply
    it seems like there's some outliers in the dataset we're just noticing now
    i wonder how ๐ŸธTTS didn't explode haha
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Edresson ๐Ÿธ: -- I'm listening to the in and out of domain test data for the 6 languages you were able to train... they sounds metallic... what can we do about this?
    1 reply
    (synthesized by @dadelani ^^^)
    nothing we can do in time for interspeech submission, but for releasing it would be good to have something less metallic
    does @dadelani have the most recent checkpoints?
    David Adelani
    @dadelani
    I have the latest checkpoint, although. I think the final checkpoint sounds better at least for Yoruba
    1 reply
    We can provide better synthesis for the camera ready.
    erogol ๐Ÿธ
    @golero:matrix.org
    [m]
    Can I find the paper somewhere to give it a read ?
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    dadelani (David Adelani): can you give eroogol a shareable link?
    1 reply
    Abdou Aziz Diop
    @abdouaziz
    Hi,@josh do you think it is still possible to associate Wolof with the project? I was able to find a data set in Wolof I think it should help
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Hi @abdouaziz ๐Ÿ‘‹
    For this current paper thereโ€™s not enough timeโ€ฆ but can you link to the data?
    We can train something for the future:)
    Abdou Aziz Diop
    @abdouaziz
    Yeah I see! Cool I hope it will happen in the future. I'm attaching the link below https://zenodo.org/record/4498861#.Yj5crS-l0Q0
    @josh-coqui:matrix.org
    can you share please if you have a stater notebook ??
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Maybe @weberjulian knows of a good notebook to start?
    erogol ๐Ÿธ
    @golero:matrix.org
    [m]
    Thanks :)
    weberjulian ๐Ÿธ
    @weberjulian:matrix.org
    [m]
    @abdouaziz josh ๐Ÿธ
    I don't train anything with notebooks but you can take inspiration from the python recipes we have for each model: https://github.com/coqui-ai/TTS/tree/main/recipes/ljspeech
    Abdou Aziz Diop
    @abdouaziz
    Thanks @weberjulian @josh for sharing
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    welcome to the room @Pogayo ๐Ÿ‘‹๐Ÿ˜„
    Abdou Aziz Diop
    @abdouaziz
    You are Welcome @Pogayo !
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Welcome to the chat room babat00nday_twitter (Baba) ๐Ÿค—
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    @Pogayo: -- the link to download Kenyan English from africanvoices.tech is broken :/
    do you have another link to it?
    Perez Ogayo
    @Pogayo
    @Josh. I have not uploaded it yet. I will be recording one in the studio today to have the final high quality version.
    Let me upload the current one
    Victor-Akinode
    @Victor-Akinode

    Hi, everyone.
    I am getting an error whenever I try running the TTS model. This is the error I get:
    Traceback (most recent call last):
    File "/home/victor/anaconda3/bin/tts", line 33, in <module>
    sys.exit(load_entry_point('TTS', 'console_scripts', 'tts')())
    File "/home/victor/TTS/TTS/bin/synthesize.py", line 268, in main
    synthesizer = Synthesizer(
    File "/home/victor/TTS/TTS/utils/synthesizer.py", line 76, in init
    self._load_tts(tts_checkpoint, tts_config_path, use_cuda)
    File "/home/victor/TTS/TTS/utils/synthesizer.py", line 115, in _load_tts
    self.tts_model = setup_tts_model(config=self.tts_config)
    File "/home/victor/TTS/TTS/tts/models/init.py", line 13, in setup_model
    model = MyModel.init_from_config(config, samples)
    File "/home/victor/TTS/TTS/tts/models/vits.py", line 1643, in init_from_config
    tokenizer, new_config = TTSTokenizer.init_from_config(config)
    File "/home/victor/TTS/TTS/tts/utils/text/tokenizer.py", line 174, in init_from_config
    characters, new_config = Graphemes().init_from_config(config)
    File "/home/victor/TTS/TTS/tts/utils/text/characters.py", line 446, in init_from_config
    Graphemes(
    File "/home/victor/TTS/TTS/tts/utils/text/characters.py", line 433, in init
    super().init(characters, punctuations, pad, eos, bos, blank, is_unique, is_sorted)
    File "/home/victor/TTS/TTS/tts/utils/text/characters.py", line 167, in init
    self._create_vocab()
    File "/home/victor/TTS/TTS/tts/utils/text/characters.py", line 261, in _create_vocab
    assert (
    AssertionError: [!] There are duplicate characters in the character set. {'(', ':', "'", '!', '.', ' ', ',', '-', ';', ')', '?'}

    Any solution to this?

    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    @Victor-Akinode: this is probably a better question for the TTS room... I'll invite you there:)
    Victor-Akinode
    @Victor-Akinode
    All right. Thank you.
    Victor-Akinode
    @Victor-Akinode
    @josh-coqui:matrix.org , please I am yet to get the invite.
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    strange-- I sent it
    one sec
    Victor-Akinode
    @Victor-Akinode
    Thank you, @josh-coqui:matrix.org
    Abdou Aziz Diop
    @abdouaziz
    Hi Here , in your experience, what should be the right (minimum) data size for TTS? And how many epochs did you train it using Coqui-ai to have a good performance on it ?
    Victor-Akinode
    @Victor-Akinode
    Hello.
    Please is there a way to run Coqui TTS on a machine with CPU without CUDA?
    josh ๐Ÿธ
    @josh-coqui:matrix.org
    [m]
    Sure:)
    TTS for inference doesnโ€™t require cuda
    This kind of question is better for the main tts room, btw
    You will get faster help there ๐Ÿ‘