You can deploy this model and start earning money today!

k-m-irfan/
Fastspeech2_HS_Flask_API

Name: Fastspeech2_HS_Flask_API
Availability: InStock

Views

No views yet

Demo: IITM-TTS Demo | This may take approximately 30 seconds to load the first time and will go idle after 15 minutes of inactivity.

Fastspeech2_HS_Flask_API

This repository contains the Flask API implementation of the Text to Speech Model developed by the Speech Lab at IIT Madras. For a comprehensive understanding of the models and inference details, please consult the original repository Fastspeech2_HS.

Setup
Installation
Run Flask server
API
Citation for the original repo

Setup

Some of the large files in this repo are uploaded using git lfs. Install latest git LFS by following the given commands:

Some of the large files in this repository have been uploaded using Git-LFS. To ensure seamless handling of these files, please install Git-LFS by executing the provided commands:

bash

curl -s https://packagecloud.io/install/repositories/github/git-lfs/script.python.sh | bash
sudo apt-get install git-lfs
git lfs install

The entire repository, including the models, has been uploaded to Hugging Face "Fastspeech2_HS_Flask_API" due to size restrictions on GitHub for Git LFS. To clone the repository from Hugging Face, please use the following command:

git clone https://huggingface.co/k-m-irfan/Fastspeech2_HS_Flask_API

Alternatively, you can download the models from the original repository Fastspeech2_HS and organize the folder structure as specified below. Skip this step if already cloned the repository from Hugging Face.

bash

models
├── hindi
│   ├── female
│   └── male
├── tamil
│   ├── female
│   └── male
.
.
.
└── marathi
    ├── female
    └── male

Installation:

Create a virtual environment and activate it:

bash

python3 -m venv tts-hs-hifigan
source tts-hs-hifigan/bin/activate

Install the required dependencies by running:

pip install -r requirements.txt

Run Flask server:

Ensure the server application is running correctly before proceeding. Use the following commands and check for any errors:

bash

python3 flask_app.py
# OR
gunicorn -w 2 -b 0.0.0.0:5000 flask_app:app --timeout 600

If the application is running without any issues, proceed to start the server using the following command:

bash start.sh

API

python

"""
This is a sample API code to send a text to the server and recieve speech
for the given text.

Supported languages: 

Assamese, Bengali, Bodo, Gujarati, Hindi, Kannada, Malayalam, Manipuri
Marathi, Odia, Punjabi, Rajasthani, Tamil, Telugu, Urdu

"""
import requests
import json
import base64

# endpoint
url = "http://localhost:5000/tts"

lang = 'hindi'
gender = 'female'
text = "सुप्रभात, आप कैसे हैं?" # hindi
# text = "സുപ്രഭാതം, സുഖമാ?" # malayalam
# text = "সুপ্ৰভাত, তুমি কেনে?" # manipuri
# text = "सुप्रभात, तुम्ही कसे आहात?" # marathi
# text = "ಶುಭೋದಯ, ನೀವು ಹೇಗಿದ್ದೀರಿ?" # kannada
# text = "बसु म्विथ्बो, बरि दिबाबो?" # bodo male yet to be added <---
# text = "Good morning, how are you?" # english
# text = "সুপ্ৰভাত, আপুনি কেমন আছে?" # assamese
# text = "காலை வணக்கம், நீங்கள் எப்படி இருக்கின்றீர்கள்?" # tamil
# text = "ସୁପ୍ରଭାତ, ଆପଣ କେମିତି ଅଛନ୍ତି?"
# text = "सुप्रभात, आप कैसे छो?" # rajasthani
# text = "శుభోదయం, మీరు ఎలా ఉన్నారు?" # telugu
# text = "সুপ্রভাত, আপনি কেমন আছেন?" # bengali
# text = "સુપ્રભાત, તમે કેમ છો?" # gujarati

payload = json.dumps(
    {
    "input": text,
    "gender": gender,
    "lang": lang,
    "alpha": 1 # to control speed
    })

headers = {'Content-Type': 'application/json'}
response = requests.request("POST", url, headers=headers, data=payload).json()

# save the received encoded audio
audio = response['audio']
file_name = "tts.wav"
wav_file = open(file_name,'wb')
decode_string = base64.b64decode(audio)
wav_file.write(decode_string)
wav_file.close()

Citation for the original repo

If you use this Fastspeech2 Model in your research or work, please consider citing:

“ COPYRIGHT 2023, Speech Technology Consortium, Bhashini, MeiTY and by Hema A Murthy & S Umesh, DEPARTMENT OF COMPUTER SCIENCE AND ENGINEERING and ELECTRICAL ENGINEERING, IIT MADRAS. ALL RIGHTS RESERVED "

Shield: