Nelson Yalta

Thank you for your collaboration. Could you add the information about this corpus on https://github.com/espnet/espnet/blob/master/egs/README.md

I am wondering about this PR, which are the problems before merging? (besides the conflicts?) Also, It will be better to prepare an organization and a list to check the...

If there is no modification in the `run.sh` file of the recipe, then the default is https://github.com/espnet/espnet/blob/f2778f798b76e102602078213a21b63aa592be70/egs2/TEMPLATE/asr1/asr.sh#L122

@mergify close

The bash code only supports fbank and raw inputs. Given the case, you can implement a frontend for raw inputs that support mfcc using torchaudio (https://pytorch.org/audio/main/generated/torchaudio.transforms.MFCC.html), and add the option...

Did you try with VCTK, or LIBRITTS vocoders?, those were trained in multispk conditions: [vctk_parallel_wavegan.v1](https://drive.google.com/open?id=1dGTu-B7an2P5sEOepLPjpOaasgaSnLpi) [vctk_parallel_wavegan.v1.long](https://drive.google.com/open?id=1qoocM-VQZpjbv5B-zVJpdraazGcPL0So) [vctk_multi_band_melgan.v2](https://drive.google.com/open?id=17EkB4hSKUEDTYEne-dNHtJT724hdivn4) [vctk_hifigan.v1](https://drive.google.com/open?id=17fu7ukS97m-8StXPc6ltW8a3hr0fsQBP) [libritts_hifigan.v1](https://drive.google.com/open?id=10jBLsjQT3LvR-3GgPZpRvWIWvpGjzDnM)

Recipe: depends on your language and target: Librispeech for reading English (clean and probably noisy environments), CommonVoice, Tedlium3, Switchboard for conversational. MLS: multilingual speech (for Spanish, Italian, and others). You...

@sw005320 Sure, I will be checking it

@iamanigeeit , If possible, add some test for `the espnet2/text/mfa_cleaners.py` and the new lines you are adding at `espnet2/text/cleaner.py`

You can use `[skip ci]` at the beginning of the commit message to avoid ci test. Also, you can run the test script on your local environment using ` ./ci/test_*.sh`...