|
# AI4Bharat-TTS Inference |
|
|
|
Text-to-Speech models trained by [AI4Bhārat](https://ai4bharat.iitm.ac.in) for 15 major languages spoken in the Indian Republic, supporting both male and female speakers. |
|
|
|
## Details |
|
|
|
### Dataset |
|
|
|
The models were trained using [the TTS dataset built by IIT-M's SMT Lab](https://www.iitm.ac.in/donlab/tts/database.php). |
|
|
|
### Languages |
|
|
|
The list of 15 languages include Indian-English, 2 Tibeto-Burman languages (Bodo & Meitei from northeast) and 12 Indic languages (4 Dravidian from South-India and 8 Indo-Aryan from Northern-India). |
|
|
|
| **Language** | **Code** | **Speakers** | **Script** | **Family** | **Native Region** | |
|
|--------------|----------|--------------|------------------------|---------------|---------------------------| |
|
| Assamese | as | male, female | Eastern-Nagari | Indo-Aryan | Assam | |
|
| Bangla | bn | male, female | Eastern-Nagari | Indo-Aryan | West-Bengal, Bangladesh | |
|
| Boro | brx | female | DevaNagari | Tibeto-Burman | Bodoland Territory | |
|
| English | en | male, female | Roman | European | -- Lingua franca -- | |
|
| Hinglish | en+hi | male, female | Code-mixed | Indo-European | | |
|
| Gujarati | gu | male, female | Gujrati | Indo-Aryan | Gujarat | |
|
| Hindi | hi | male, female | DevaNagari | Indo-Aryan | Hindi Belt | |
|
| Kannada | kn | male, female | Kannada | Dravidian | Karnataka | |
|
| Malayalam | ml | male, female | Malayalam | Dravidian | Kerala | |
|
| Manipuri | mni | male, female | Meetei, Eastern-Nagari | Tibeto-Burman | Imphal valley (Manipur) | |
|
| Marathi | mr | male, female | DevaNagari | Indo-Aryan | Maharashtra | |
|
| Oriya | or | male, female | Odia | Indo-Aryan | Odisha | |
|
| Panjabi | pa | male, female | Gurumukhi | Indo-Aryan | Eastern-Punjab | |
|
| Rajasthani | raj | male, female | DevaNagari | Indo-Aryan | Rajasthan | |
|
| Tamil | ta | male, female | Tamil | Dravidian | Tamil Nadu | |
|
| Telugu | te | male, female | Telugu | Dravidian | Andhra Pradesh, Telangana | |
|
|
|
## Usage |
|
|
|
### Pre-requisites |
|
|
|
1. Python 3.9+ |
|
2. If Linux, install the following dependencies: |
|
``` |
|
cd inference |
|
sudo apt-get install libsndfile1-dev ffmpeg enchant |
|
``` |
|
(For any other OS, we wish you best of luck) |
|
3. `pip install -r requirements-ml.txt requirements-utils.txt` |
|
4. [Download the models from here](https://github.com/AI4Bharat/Indic-TTS/releases), place them inside a new folder named `checkpoints` and unzip them. |
|
|
|
### Running inference |
|
|
|
Check `sample.py` for usage. |
|
|
|
### Hosting REST API server |
|
|
|
``` |
|
pip install -r requirements-server.txt |
|
uvicorn server:api |
|
``` |
|
|