This example contains code used to train a [Fastspeech2](https://arxiv.org/abs/2006.04558) model with [CSMSC](https://www.data-baker.com/open_source.html), [LJSpeech-1.1](https://keithito.com/LJ-Speech-Dataset/), [AISHELL3](http://www.aishelltech.com/aishell_3) and [VCTK](https://datashare.ed.ac.uk/handle/10283/3443) datasets.
Or train your MFA model reference to [mfa example](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/other/mfa) (use MFA1.x now) of our repo.
You can choose a range of stages you want to run, or set `stage` equal to `stop-stage` to use only one stage, for example, running the following command will only preprocess the dataset.
The dataset is split into 3 parts, namely `train`, `dev`, and` test`, each of which contains a `norm` and `raw` subfolder. The raw folder contains speech, pitch and energy features of each utterance, while the norm folder contains normalized ones. The statistics used to normalize features are computed from the training set, which is located in `dump/train/*_stats.npy`.
Also, there is a `metadata.jsonl` in each subfolder. It is a table-like file that contains phones, text_lengths, speech_lengths, durations, the path of speech features, the path of pitch features, a path of energy features, speaker, and id of each utterance.
We have **added module speaker classifier** with reference to [Learning to Speak Fluently in a Foreign Language: Multilingual Speech Synthesis and Cross-Language Voice Cloning](https://arxiv.org/pdf/1907.04448.pdf). The main parameter configuration: `config["model"]["enable_speaker_classifier"]`, `config["model"]["hidden_sc_dim"]` and `config["updater"]["spk_loss_scale"]` in `conf/default.yaml`. The current experimental results show that this module can decouple text information and speaker information, and more experiments are still being sorted out. This module is currently not enabled by default, if you are interested, you can try it yourself.
Download the pretrained parallel wavegan model from [pwg_aishell3_ckpt_0.5.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/pwgan/pwg_aishell3_ckpt_0.5.zip) and unzip it.
When speaker is `174` (csmsc), use csmsc's vocoder is better than aishell3's, we recommend that you use [hifigan_csmsc_ckpt_0.1.1.zip](https://paddlespeech.bj.bcebos.com/Parakeet/released_models/hifigan/hifigan_csmsc_ckpt_0.1.1.zip), please check `stage 2` of `synthesize_e2e.sh`.
But if speaker is `175` (ljspeech), we **don't** recommend you to use ljspeech's vocoder, because ljspeech's vocoders are trained on sample rate 22.05kHz, but this acoustic model is trained on sample rate 24kHz, you can use csmsc's vocoder also, because ljspeech and csmsc are both female speakers.
For speakers in aishell3 and vctk, we recommend you use aishell3 or vctk's vocoders, because ljspeech and csmsc are both female speakers, there vocoders may not perform well for male speakers in aishell3 and vctk, you can check speaker name and spk_id in `dump/speaker_id_map.txt` and check speakers' information ( Age / Gender / Accents / region, etc ) in [this issue](https://github.com/PaddlePaddle/PaddleSpeech/issues/1620) and choose the `spk_id` you want.
--voc_stat VOC_STAT mean and standard deviation used to normalize
spectrogram when training voc.
--lang LANG Choose model language. zh or en or mix
--inference_dir INFERENCE_DIR
dir to save inference models
--ngpu NGPU if ngpu == 0, use cpu.
--text TEXT text to synthesize, a 'utt_id sentence' pair per line.
--output_dir OUTPUT_DIR
output dir.
```
1.`--am` is acoustic model type with the format {model_name}_{dataset}
2.`--am_config`, `--am_ckpt`, `--am_stat`, `--phones_dict``--speaker_dict` are arguments for acoustic model, which correspond to the 5 files in the fastspeech2 pretrained model.
3.`--voc` is vocoder type with the format {model_name}_{dataset}
4.`--voc_config`, `--voc_ckpt`, `--voc_stat` are arguments for vocoder, which correspond to the 3 files in the parallel wavegan pretrained model.
5.`--lang` is the model language, which can be `zh` or `en` or `mix`.
6.`--test_metadata` should be the metadata file in the normalized subfolder of `test` in the `dump` folder.
7.`--text` is the text file, which contains sentences to synthesize.
8.`--output_dir` is the directory to save synthesized audio files.
9.`--ngpu` is the number of gpus to use, if ngpu == 0, use cpu.
## Pretrained Model
Pretrained FastSpeech2 model with no silence in the edge of audios:
