# tacotron-2-support
- h
hecko
12/21/2022, 10:39 AM14.4% word error rate - e
Edward
12/21/2022, 10:39 AMOn what model? - h
hecko
12/21/2022, 10:39 AMbut we'll see how much that matters - h
hecko
12/21/2022, 10:39 AMpresumably large - e
Edward
12/21/2022, 10:39 AMOh - e
Edward
12/21/2022, 10:40 AMI tested it once and it gave me a perfect transcription - e
Edward
12/21/2022, 10:40 AMNeeds testing - e
Edward
12/21/2022, 10:40 AMSo i still need help with this error... - h
hecko
12/21/2022, 10:41 AMthis is how to solve it - e
Edward
12/21/2022, 10:42 AMIn what file? - e
Edward
12/21/2022, 11:01 AMCould I use the base model of my voice to train another model of another voice like a news presenter and use that as the public Romanian base model? @hecko - h
hecko
12/21/2022, 11:02 AMnot file, the code of the training cell - h
hecko
12/21/2022, 11:03 AMsure - e
Edward
12/21/2022, 11:03 AMOkay good - e
Edward
12/21/2022, 12:03 PMi'm not training, i'm trying to synthetise using https://colab.research.google.com/github/justinjohn0306/FakeYou-Tacotron2-Notebook/blob/main/Latest_Tacotron2_and_HiFi_GAN.ipynb - h
hecko
12/21/2022, 12:03 PMoh uhh - h
hecko
12/21/2022, 12:03 PMidk about multilingual support in that notebook - h
hecko
12/21/2022, 12:04 PMi know the pipeline notebook supports languages but i don't think it has romanian - e
Edward
12/21/2022, 12:04 PM@Justin can you help ? - j
Justin
12/21/2022, 12:47 PMIt's meant only for English - j
Justin
12/21/2022, 12:48 PMUse pipeline training nb instead - e
Edward
12/21/2022, 12:48 PMit's already trained - e
Edward
12/21/2022, 12:48 PMi just want synthesis - j
Justin
12/21/2022, 12:49 PMokay - j
Justin
12/21/2022, 12:49 PMit won't support the symbol set for the Romanian language - j
Justin
12/21/2022, 12:49 PMyou can synthesize it anyways - j
Justin
12/21/2022, 12:49 PMI'll DM you - e
Edward
12/21/2022, 12:49 PMokay thanks - p
PixPrucer
12/21/2022, 6:29 PM@hecko yup yup yup - h
hecko
12/21/2022, 6:33 PM🎉