# ai4bharat/indic-tts architecture

> Architecture diagram of the GitHub repository ai4bharat/indic-tts, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-02). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/ai4bharat/indic-tts
- Repository: https://github.com/ai4bharat/indic-tts

## Overview

AI4Bharat Indic-TTS provides training and inference for Indian-language speech synthesis, with published models for 13 languages. The documented workflow formats and analyzes the IndicTTS dataset, configures FastPitch acoustic-model and HiFi-GAN vocoder training, then serves text-to-speech through an API or Triton. The inference engine normalizes and transliterates text, segments it, synthesizes audio, and postprocesses the result. The graph also preserves documented preprocessing, training, and deployment capabilities; unsampled training and socket-proxy implementation details are represented only where the tree supports a path. Edges cite the README or shown caller code.

Main entry points: [`inference/src/inference.py`](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/inference.py) (TTS engine), [`inference/server.py`](https://github.com/ai4bharat/indic-tts/blob/master/inference/server.py) (HTTP API), [`configs`](https://github.com/ai4bharat/indic-tts/tree/master/configs) (Training configs), [`inference/src/utils/text.py`](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/text.py) (Text normalization)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_data["Dataset preparation"]
  node_dataset[("IndicTTS dataset")]
  node_format["Dataset formatter"]
  node_analyze["Dataset analysis"]
end

subgraph group_training["Model training"]
  node_acoustic["FastPitch training<br/>[main.py]"]
  node_vocoder["HiFi-GAN training<br/>[vocoder.py]"]
  node_config["Training configs"]
end

subgraph group_inference["Speech inference"]
  node_engine["TTS engine<br/>[inference.py]"]
  node_requests["Request models<br/>[request.py]"]
  node_responses["Audio responses<br/>[response.py]"]
  node_normalizer["Text normalization<br/>[text.py]"]
  node_translator["Text translation<br/>[translator.py]"]
  node_paragraphs["Paragraph splitting"]
  node_postprocessor["Audio postprocessing<br/>[postprocessor.py]"]
  node_vad["Voice activity detection<br/>[vad.py]"]
  node_denoiser["Audio denoiser<br/>[denoiser.py]"]
end

subgraph group_serving["Serving interfaces"]
  node_api["HTTP API<br/>[server.py]"]
  node_sample["Inference sample<br/>[sample.py]"]
  node_triton["Triton model<br/>[model.py]"]
  node_tritonclient["Triton client<br/>[client.py]"]
  node_socket["Socket proxy<br/>[te_endpoint.py]"]
  node_webclient["Browser client<br/>[main.js]"]
end

node_user(("TTS caller"))

node_dataset -->|"formats"| node_format
node_dataset -->|"analyzes"| node_analyze
node_config -->|"configures"| node_acoustic
node_config -->|"configures"| node_vocoder
node_user -->|"sends request"| node_api
node_api -->|"calls inference"| node_engine
node_api -->|"accepts"| node_requests
node_engine -->|"reads"| node_requests
node_engine -->|"constructs"| node_responses
node_engine -->|"normalizes"| node_normalizer
node_normalizer -->|"translates text"| node_translator
node_engine -->|"splits text"| node_paragraphs
node_engine -->|"processes audio"| node_postprocessor
node_postprocessor -->|"trims silence"| node_vad
node_engine -->|"denoises audio"| node_denoiser
node_user -.->|"runs example"| node_sample
node_user -.->|"uses client"| node_tritonclient
node_triton -->|"calls inference"| node_engine
node_webclient -.->|"connects"| node_socket

click node_format "https://github.com/ai4bharat/indic-tts/blob/master/preprocessing/FormatDataset-IndicTTS.ipynb"
click node_analyze "https://github.com/ai4bharat/indic-tts/blob/master/preprocessing/AnalyzeDataset-IndicTTS.ipynb"
click node_acoustic "https://github.com/ai4bharat/indic-tts/blob/master/main.py"
click node_vocoder "https://github.com/ai4bharat/indic-tts/blob/master/vocoder.py"
click node_config "https://github.com/ai4bharat/indic-tts/tree/master/configs"
click node_api "https://github.com/ai4bharat/indic-tts/blob/master/inference/server.py"
click node_sample "https://github.com/ai4bharat/indic-tts/blob/master/inference/sample.py"
click node_triton "https://github.com/ai4bharat/indic-tts/blob/master/inference/triton_server/triton_repo/tts/1/model.py"
click node_tritonclient "https://github.com/ai4bharat/indic-tts/blob/master/inference/triton_server/client.py"
click node_socket "https://github.com/ai4bharat/indic-tts/blob/master/inference/socket_proxy/te_endpoint.py"
click node_webclient "https://github.com/ai4bharat/indic-tts/blob/master/inference/socket_proxy/front_end/main.js"
click node_engine "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/inference.py"
click node_requests "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/models/request.py"
click node_responses "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/models/response.py"
click node_normalizer "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/text.py"
click node_translator "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/translator.py"
click node_paragraphs "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/paragraph_handler.py"
click node_postprocessor "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/postprocessor.py"
click node_vad "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/vad.py"
click node_denoiser "https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/denoiser.py"

classDef toneNeutral fill:#f8fafc,stroke:#334155,stroke-width:1.5px,color:#0f172a
classDef toneBlue fill:#dbeafe,stroke:#2563eb,stroke-width:1.5px,color:#172554
classDef toneAmber fill:#fef3c7,stroke:#d97706,stroke-width:1.5px,color:#78350f
classDef toneMint fill:#dcfce7,stroke:#16a34a,stroke-width:1.5px,color:#14532d
classDef toneRose fill:#ffe4e6,stroke:#e11d48,stroke-width:1.5px,color:#881337
classDef toneIndigo fill:#e0e7ff,stroke:#4f46e5,stroke-width:1.5px,color:#312e81
classDef toneTeal fill:#ccfbf1,stroke:#0f766e,stroke-width:1.5px,color:#134e4a
class node_dataset,node_format,node_analyze toneBlue
class node_acoustic,node_vocoder,node_config toneAmber
class node_engine,node_requests,node_responses,node_normalizer,node_translator,node_paragraphs,node_postprocessor,node_vad,node_denoiser toneMint
class node_api,node_sample,node_triton,node_tritonclient,node_socket,node_webclient toneRose
class node_user toneIndigo
```

## Components

### Dataset preparation

- **IndicTTS dataset**
- [Dataset formatter](https://github.com/ai4bharat/indic-tts/blob/master/preprocessing/FormatDataset-IndicTTS.ipynb): `preprocessing/FormatDataset-IndicTTS.ipynb`
- [Dataset analysis](https://github.com/ai4bharat/indic-tts/blob/master/preprocessing/AnalyzeDataset-IndicTTS.ipynb): `preprocessing/AnalyzeDataset-IndicTTS.ipynb`

### Model training

- [FastPitch training](https://github.com/ai4bharat/indic-tts/blob/master/main.py): `main.py`
- [HiFi-GAN training](https://github.com/ai4bharat/indic-tts/blob/master/vocoder.py): `vocoder.py`
- [Training configs](https://github.com/ai4bharat/indic-tts/tree/master/configs): `configs`

### Speech inference

- [TTS engine](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/inference.py): `inference/src/inference.py`
- [Request models](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/models/request.py): `inference/src/models/request.py`
- [Audio responses](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/models/response.py): `inference/src/models/response.py`
- [Text normalization](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/text.py): `inference/src/utils/text.py`
- [Text translation](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/translator.py): `inference/src/utils/translator.py`
- [Paragraph splitting](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/utils/paragraph_handler.py): `inference/src/utils/paragraph_handler.py`
- [Audio postprocessing](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/postprocessor.py): `inference/src/postprocessor/postprocessor.py`
- [Voice activity detection](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/vad.py): `inference/src/postprocessor/vad.py`
- [Audio denoiser](https://github.com/ai4bharat/indic-tts/blob/master/inference/src/postprocessor/denoiser.py): `inference/src/postprocessor/denoiser.py`

### Serving interfaces

- [HTTP API](https://github.com/ai4bharat/indic-tts/blob/master/inference/server.py): `inference/server.py`
- [Inference sample](https://github.com/ai4bharat/indic-tts/blob/master/inference/sample.py): `inference/sample.py`
- [Triton model](https://github.com/ai4bharat/indic-tts/blob/master/inference/triton_server/triton_repo/tts/1/model.py): `inference/triton_server/triton_repo/tts/1/model.py`
- [Triton client](https://github.com/ai4bharat/indic-tts/blob/master/inference/triton_server/client.py): `inference/triton_server/client.py`
- [Socket proxy](https://github.com/ai4bharat/indic-tts/blob/master/inference/socket_proxy/te_endpoint.py): `inference/socket_proxy/te_endpoint.py`
- [Browser client](https://github.com/ai4bharat/indic-tts/blob/master/inference/socket_proxy/front_end/main.js): `inference/socket_proxy/front_end/main.js`

### Other components

- **TTS caller**

## Connections

- IndicTTS dataset → Dataset formatter: formats (evidence: `README.md`)
- IndicTTS dataset → Dataset analysis: analyzes (evidence: `README.md`)
- Training configs → FastPitch training: configures (evidence: `README.md`)
- Training configs → HiFi-GAN training: configures (evidence: `README.md`)
- TTS caller → HTTP API: sends request (evidence: `inference/server.py`)
- HTTP API → TTS engine: calls inference (evidence: `inference/server.py`)
- HTTP API → Request models: accepts (evidence: `inference/server.py`)
- TTS engine → Request models: reads (evidence: `inference/src/inference.py`)
- TTS engine → Audio responses: constructs (evidence: `inference/src/inference.py`)
- TTS engine → Text normalization: normalizes (evidence: `inference/src/inference.py`)
- Text normalization → Text translation: translates text (evidence: `inference/src/utils/text.py`)
- TTS engine → Paragraph splitting: splits text (evidence: `inference/src/inference.py`)
- TTS engine → Audio postprocessing: processes audio (evidence: `inference/src/inference.py`)
- Audio postprocessing → Voice activity detection: trims silence (evidence: `inference/src/postprocessor/postprocessor.py`)
- TTS engine → Audio denoiser: denoises audio (evidence: `inference/src/inference.py`)
- TTS caller → Inference sample: runs example
- TTS caller → Triton client: uses client
- Triton model → TTS engine: calls inference (evidence: `inference/triton_server/triton_repo/tts/1/model.py`)
- Browser client → Socket proxy: connects
