coqui_tts_korea

Coqui_TTS_KOREA

Korean TTS Using Coqui TTS (Glowts and Multiband Melgan) -Korean TTS

• colab

prerain with kss data
FINETUNE HALFLIFE Scientist Data

Result

• INPUT TEXT
  • "God is not interested in our mathematics problems. God is just empirically integrated."
• output

GlowTTS-MBMELGAN.MOV

HALFLIFE_MBMELGAN.MOV

Train Detail

• Glowts

  • Trained with KSS DATA 190000 Step
  • Train iPynb file: Coqui_train_glowts.ipynb
  • Google Drive Link: https://drive.google.com/drive/folders/1Quloabjkammw6mfbccsmmmmmmmmmmbbcW

• Multiband-Melgan

  • Trained with Korea Concat Data (KSS, Zeroth and Pansori-Tedxkr) 150000 Step
  • Train iPynb file: Coqui_train_mbmelgan.ipynb
  • Google Drive Link: https://drive.google.com/drive/folders/1Folcojx47j_alnw28Rzkr62iowQHY6TE

• HALFLIFE FINETUNED GLOWTTS

  • Trained with KSS DATA 190000 Step + Halflife 90000 STEP
  • Train iPynb file: Halflife_finetune_glowts.ipynb
  • Google Drive Link: https://drive.google.com/drive/folders/1rubvjsdkz_HNP3XJ8MCOCWG3KBMT4R?usp=sharing

• HALFLIFE FINETUNED MULTIBAND-MELGAN

  • Trained with KOREA CONCAT DATA
  • Train iPynb File: Halflife_finetune_mbmelgan.ipynb
  • Google Drive Link: https://drive.google.com/drive/folders/15eaW8jsioisipqa03vomoh-pacguc?usp=sharing

Dataset

• KSS Dataset
• Zeroth Korean
• Pansori-Tedxkr
• HALF_LIFE_DATASET

Required Environment to Run

!p ip install TTS
!p ip install jamo
!p ip install torchaudio == 0.9 . 0
!p ip install gdown
!c onda install - c conda - forge kaggle - y
!p ip install librosa

ACKNOWEDGEMENT and References

• Coqui TTS
• Tensorflowts
• Glow TTS
• Multi-Band Melgan
• FastSpeech 2
• SPEECH-JAPANESE-KOREAN-VIETNAMESE
• openslr
• HALF_LIFE_DATASET
• KSS Dataset
• Zeroth Korean
• Pansori-Tedxkr
• Fine-Tuning with a Small Dataset
• Change SIRI into IU voice
• Implementing TTS (voice synthesis) using artificial intelligence Deep Voice_Son Seok -hee Anchor
• SCE-TTS: Create TTS in my voice
• HuggingFace_fastspeech2_kss
• HuggingFace_Tensorflowts
• jamo
• Korean.py
• FreeConvert