Government Of India
A-
A
A+
Bhashini - Fastspeech2 Model using (HS)
Text-to-speech models trained using FastPitch and HiFi-GAN vocoder, separately for each language. Supports both 'female' and 'male' voices.
About Model
This repository contains a Fastspeech2 Model for 16 Indian languages (male and female both) implemented using the Hybrid Segmentation (HS) for speech synthesis. The model is capable of generating mel-spectrograms from text inputs and can be used to synthesize speech.
Fs2 is composed of 6 feed-forward Transformer blocks with multi-head self-attention and 1D convolution on both phoneme encoder and mel-spectrogram decoder. In each feed-forward Transformer, the hidden size of multi-head attention is set to 256 and the number of head is set to 2. The kernel size of 1D convolution in the two-layer convolution network is set to 9 and 1, and the input/output size of the number of channels in the first and the second layer is 256/1024 and 1024/256. The duration predictor and variance adaptor, which are composed of stacks of several convolution networks and the final linear projection layer. The convolution layers of the duration predictor and variance adaptor are set to 2 and 5, the kernel size is set to 3, the input/output size of all layers is 256/256, and the dropout rate is set to 0.5.
Bhashini - Fastspeech2 Model using (HS)
Metadata
MIT
SMT Lab IIT Madras
Speech Synthesis (TTS) Model
Other
Open
Sector Agnostic
01/05/25 06:47:30
286.72 MB
Activity Overview
0
80 
1,233 
286.72 MB
Tags
- Multilingual
- NLP
- Text Processing
- Transformer
- Text to Speech
- Language Detection
License Control
MIT
Version Control
Version 2(286.72 MB)
- admin·1 year(s) ago
- chevron_right
assamese
- chevron_right
female
- chevron_right
male
- chevron_right
bengali
- chevron_right
bodo
- chevron_right
charmap
- chevron_right
english
.gitattributes
api.py
app.py
environment.yml
get_phone_mapped_python.py
- more_horiz 25 more
Version 1(7.06 KB)
- admin·1 year(s) ago
No File(s) Found!
More Models from Digital India BHASHINI Division
SANTHAM-Gemma3-4B-SH-Seg-Poetry-Finetuned
SANTHAM-Gemma3-4B-SH-Seg-Poetry-Finetuned is a model designed to translate Sanskrit into Tamil specialized on Segmented text obtained using Sanskrit Heritage segmenter.
translation
poetry
santham
Segmened
language:tam
language:san
- 0
- 0
- 115.62 MB
49
Updated 20 day(s) ago
DIGITAL INDIA BHASHINI DIVISION
SANTHAM-Gemma3-4B-Finetuned
SANTHAM-Gemma3-4B-Finetuned is a Sanskrit → Tamil translation model built on the Gemma 3 (4B) architecture.
It is trained on a parallel corpus developed as part of the Sanskrit Knowledge Accessor project, enabling it to capture linguistic nuances and generate fluent Tamil translations from classical Sanskrit inputs.
translation
language:san
language:tam
santham
- 0
- 2
- 2.08 GB
- 70
Updated 20 day(s) ago
DIGITAL INDIA BHASHINI DIVISION
SANTHAM-Gemma3-4B-Anvaya-Poetry-Finetuned
SANTHAM-Gemma3-4B-Anvaya-Potery-Finetuned is a model designed to translate Sanskrit into Tamil specialized on Anvaya translation in Poetry.
poetry
santham
anvaya
language:tam
language:san
translation
- 0
- 2
- 2.09 GB
- 43
Updated 20 day(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-URDU
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
spring_lab
Data2vec_aqc
low-resource-language
SSL_finetunning
ssl
urdu
IITM
- 0
- 0
- 3.52 GB
- 66
Updated 1 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-TELUGU
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
spring_lab
low-resource-language
SSL_finetunning
Data2vec_aqc
IITM
telugu
ssl
- 0
- 0
- 3.52 GB
- 56
Updated 1 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-TAMIL
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
low-resource-language
SSL_finetunning
Data2vec_aqc
spring_lab
IITM
tamil
ssl
- 0
- 1
- 3.52 GB
- 52
Updated 1 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-BENGALI
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
Data2vec_aqc
IITM
spring_lab
ssl
low-resource-languages
SSL_finetunning
bengali
- 0
- 0
- 3.52 GB
- 93
Updated 2 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-BODO
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
IITM
spring_lab
SSL_finetunning
low-resource-language
BODO
Data2vec_aqc
ssl
- 0
- 0
- 3.52 GB
- 122
Updated 2 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-BHOJPURI
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
SSL_finetunning
Data2vec_aqc
spring_lab
IITM
ssl
Bhojpuri
low-resource-language
- 0
- 4
- 3.52 GB
- 113
Updated 2 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
SPRING-INX-DATA2VEC-AQC-MALAYALAM
Automatic Speech Recognition (ASR) model for speech recognition, processing audio and transcribing spoken content into text.
SSL_finetunning
ssl
malayalam
IITM
spring_lab
Data2vec_aqc
low-resource-language
- 0
- 1
- 3.52 GB
- 119
Updated 2 month(s) ago
DIGITAL INDIA BHASHINI DIVISION
© 2026 - Copyright AIKosh. All rights reserved. This portal is developed by National e-Governance Division for AIKosh mission.