Starstruk Logo MarkStarstruk

r0.1.0

PaddlePaddle/PaddleSpeechr0.1.0Dec 23, 2021by Jackwaterveg

AI Summary

The initial release of PaddleSpeech, establishing a unified platform for speech tasks including ASR, TTS, Speech Translation, and Audio Classification with a robust CLI.

Key Highlights

  • Unified CLI for ASR, TTS, ST, and Audio Classification.
  • Introduced FastSpeech2-Conformer and multiple vocoders (HiFiGAN, Style MelGAN, Multi Band MelGAN).
  • Added DeepSpeech2 online model and Transformer LM.
  • Integrated FastSpeech2 Voice Cloning using GE2E.
  • Added 9 interactive demos (subtitles, metaverse, etc.).

New Features

  • Easy installation via `pip install paddlespeech`
  • Audio classification on ESC-50 and custom datasets
  • ST-MTL and FAT-ST-MTL models
  • DeepSpeech2 CTC LM decoder
  • Improved DeepSpeech2 online model

Full Release Notes

# Features
## CLI : New Feature
  -  Easy install by pip `pip install paddlespeech`
  -  CLI to quick explore ASR, TTS, audio classification, speech translation and punctuation restoration.
## ASR
  - Join CTC LM decoder
     - [paper link](https://www.merl.com/publications/docs/TR2017-190.pdf)
  - Transformer LM model 
  - Improve  DeepSpeech2 online model
  - Refactor some configs
## TTS
  -  Merge Parakeet into PaddleSpeech
  -  Add FastSpeech2-Conformer
      - paper link: [fastspeech2](https://arxiv.org/abs/2006.04558) 、[conformer](https://arxiv.org/abs/2005.08100)
      - [example link](https://github.com/PaddlePaddle/PaddleSpeech/blob/develop/examples/csmsc/tts3/conf/conformer.yaml)
  -  Add Multi Band MelGAN
      - [paper link](https://arxiv.org/abs/2005.05106)
      - [example link](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc3)
  -  Add HiFiGAN
      - [paper link](https://arxiv.org/abs/2010.05646) 
      - [example link](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc5)
  -  Add Style MelGAN
      - [paper link](https://arxiv.org/abs/2011.01557) 
      - [example link](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/csmsc/voc4)
  -  Add FastSpeech2 Voice Cloning with GE2E (SV2TTS)
      - [paper link](https://arxiv.org/pdf/1806.04558.pdf)
      - [example link](https://github.com/PaddlePaddle/PaddleSpeech/tree/develop/examples/aishell3/vc1)
## CLS
  -  Add audio classification example on ESC-50 and custom dataset.
  -  Add audio tagging demo based on PANNs and Audioset labels.
## ST
  -  ST-MTL
  -  FAT-ST-MTL

## Docs
  -   Add quick start
  -  Add read the doc
  -  Improve installation documentation
  -  Add README for each example
## Demos
  -  Audio_tagging
  -  Automatic_video_subtitiles
  -  Metaverse
  -  Punctuation_restoration
  -  Speech_recognition
  -  Speech_translation
  -  Story_talker
  -  Style_fs2
  -  Text_to_speech
## Others
  -  Update released models and results

# Acknowledgements
@zh794390558   @KPatr1ck  @Jackwaterveg  @yt605155624 @Mingxue-Xu  @grasswolfs  @jerryuhoo