| from os import getenv | |
| from huggingface_hub import hf_hub_download | |
| from json import load as json_load , dump as json_dump | |
| from torch import device as Device | |
| from torch.cuda import is_available as cuda_is_available | |
| from TTS.utils.synthesizer import Synthesizer | |
| lang_conf = { | |
| "as": "Assamese - অসমীয়া", | |
| "bn": "Bangla - বাংলা", | |
| "brx": "Boro - बड़ो", | |
| "en": "English (Indian accent)", | |
| "en+hi": "English+Hindi (Hinglish code-mixed)", | |
| "gu": "Gujarati - ગુજરાતી", | |
| "hi": "Hindi - हिंदी", | |
| "kn": "Kannada - ಕನ್ನಡ", | |
| "ml": "Malayalam - മലയാളം", | |
| "mni": "Manipuri - মিতৈলোন", | |
| "mr": "Marathi - मराठी", | |
| "or": "Oriya - ଓଡ଼ିଆ", | |
| "pa": "Panjabi - ਪੰਜਾਬੀ", | |
| "raj": "Rajasthani - राजस्थानी", | |
| "ta": "Tamil - தமிழ்", | |
| "te": "Telugu - తెలుగు" | |
| } | |
| class Indic_TTS: | |
| def __init__(self,lang,device): | |
| model_id = "shethjenil/INDIC_TTS" | |
| model_path = hf_hub_download(model_id, lang+"_fastpitch_best_model.pth") | |
| vocoder_path = hf_hub_download(model_id, lang+"_hifigan_best_model.pth") | |
| vocoder_config_path = hf_hub_download(model_id, lang+"_hifigan_config.json") | |
| config_path = hf_hub_download(model_id, lang+"_fastpitch_config.json") | |
| speaker_path = hf_hub_download(model_id, lang+"_fastpitch_speakers.pth") | |
| conf = json_load(open(config_path)) | |
| conf['speakers_file'] = conf['model_args']['speakers_file'] = speaker_path | |
| json_dump(conf, open(config_path, 'w')) | |
| self.synthesizer = Synthesizer(model_path,config_path,vocoder_checkpoint=vocoder_path,vocoder_config=vocoder_config_path,use_cuda=device.type == "cuda") | |
| self.speakers = self.synthesizer.tts_model.speaker_manager.speaker_names | |
| def text2speech(self,text:str,speaker:str): | |
| self.synthesizer.save_wav(self.synthesizer.tts(text,speaker),"output.wav") | |
| return "output.wav" | |
| indic_tts_lang = getenv("indic_tts_lang") | |
| tts_lang_name = lang_conf[indic_tts_lang] | |
| tts_model = Indic_TTS(indic_tts_lang,Device("cuda" if cuda_is_available() else "cpu")) | |
| import gradio as gr | |
| gr.Interface(tts_model.text2speech,[gr.Textbox(label="Enter Text"),gr.Dropdown(tts_model.speakers, label="speaker"),],gr.Audio(type="filepath", label="Speech"),title=f"{tts_lang_name} TTS").launch() | |