# iszhanjiawei/indextts2 architecture

> Architecture diagram of the GitHub repository iszhanjiawei/indextts2, generated by GitDiagram from its file tree, README and a sample of source files (last updated 2026-10-03). It is an AI-made overview: check details against the source.

- Interactive diagram: https://gitdiagram.com/iszhanjiawei/indextts2
- Repository: https://github.com/iszhanjiawei/indextts2

## Overview

IndexTTS2 is a zero-shot expressive text-to-speech system. A user supplies text and a speaker-reference recording, optionally adding an emotion reference, emotion scores, or text emotion guidance; inference generates speech with either controlled or freely generated duration. The README documents these modes, and the source index connects the inference entry point to the main model and audio-generation modules. Several internal synthesis stages are represented by their implementation directories because their source was not sampled; those relationships are not inferred beyond documented capabilities and indexed references.

Main entry points: [`indextts/infer_indextts2.py`](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/infer_indextts2.py) (Inference API), [`indextts/gpt/model_v2.py`](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/model_v2.py) (Autoregressive model), [`webui.py`](https://github.com/iszhanjiawei/indextts2/blob/main/webui.py) (Web interface), [`indextts/s2mel/modules/bigvgan/bigvgan.py`](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/s2mel/modules/bigvgan/bigvgan.py) (BigVGAN vocoder)

## Diagram (Mermaid)

```mermaid
flowchart TD

subgraph group_access["User access"]
  node_webui["Web interface<br/>[webui.py]"]
  node_api["Inference API<br/>[infer_indextts2.py]"]
end

subgraph group_conditioning["Prompt conditioning"]
  node_text["Text frontend<br/>[front.py]"]
  node_features["Audio features"]
  node_speaker["Speaker encoding<br/>[DTDNN.py]"]
  node_emotion["Emotion guidance<br/>[infer_indextts2.py]"]
end

subgraph group_generation["Speech generation"]
  node_ar["Autoregressive model<br/>[model_v2.py]"]
  node_conformer["Conformer encoder"]
  node_perceiver["Perceiver<br/>[perceiver.py]"]
  node_duration["Duration adaptation<br/>[model_v2.py]"]
end

subgraph group_audio["Audio synthesis"]
  node_mel["Mel generation"]
  node_codec["Codec components"]
  node_vocoder["BigVGAN vocoder<br/>[bigvgan.py]"]
end

node_user(("TTS user"))
node_wave["Generated audio"]

node_user -->|"submits text"| node_webui
node_user -->|"provides prompts"| node_api
node_webui -->|"invokes inference"| node_api
node_api -->|"uses frontend"| node_text
node_api -->|"uses extractors"| node_features
node_api -->|"uses encoder"| node_speaker
node_api -->|"accepts guidance"| node_emotion
node_api -->|"uses model"| node_ar
node_ar -->|"uses encoder"| node_conformer
node_ar -->|"uses perceiver"| node_perceiver
node_ar -->|"controls duration"| node_duration
node_api -->|"uses synthesis modules"| node_mel
node_api -->|"uses codec utilities"| node_codec
node_api -->|"uses vocoder"| node_vocoder
node_api -->|"writes speech"| node_wave
node_vocoder -.->|"synthesizes waveform"| node_wave

click node_webui "https://github.com/iszhanjiawei/indextts2/blob/main/webui.py"
click node_api "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/infer_indextts2.py"
click node_text "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/utils/front.py"
click node_features "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/utils/feature_extractors.py"
click node_speaker "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/s2mel/modules/campplus/DTDNN.py"
click node_emotion "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/infer_indextts2.py"
click node_ar "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/model_v2.py"
click node_conformer "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/conformer_encoder.py"
click node_perceiver "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/perceiver.py"
click node_duration "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/model_v2.py"
click node_mel "https://github.com/iszhanjiawei/indextts2/tree/main/indextts/s2mel/modules"
click node_codec "https://github.com/iszhanjiawei/indextts2/tree/main/indextts/utils/maskgct/models/codec"
click node_vocoder "https://github.com/iszhanjiawei/indextts2/blob/main/indextts/s2mel/modules/bigvgan/bigvgan.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_webui,node_api,node_user toneBlue
class node_text,node_features,node_speaker,node_emotion toneAmber
class node_ar,node_conformer,node_perceiver,node_duration toneMint
class node_mel,node_codec,node_vocoder toneRose
class node_wave toneIndigo
```

## Components

### User access

- [Web interface](https://github.com/iszhanjiawei/indextts2/blob/main/webui.py): `webui.py`
- [Inference API](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/infer_indextts2.py): `indextts/infer_indextts2.py`

### Prompt conditioning

- [Text frontend](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/utils/front.py): `indextts/utils/front.py`
- [Audio features](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/utils/feature_extractors.py): `indextts/utils/feature_extractors.py`
- [Speaker encoding](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/s2mel/modules/campplus/DTDNN.py): `indextts/s2mel/modules/campplus/DTDNN.py`
- [Emotion guidance](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/infer_indextts2.py): `indextts/infer_indextts2.py`

### Speech generation

- [Autoregressive model](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/model_v2.py): `indextts/gpt/model_v2.py`
- [Conformer encoder](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/conformer_encoder.py): `indextts/gpt/conformer_encoder.py`
- [Perceiver](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/perceiver.py): `indextts/gpt/perceiver.py`
- [Duration adaptation](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/gpt/model_v2.py): `indextts/gpt/model_v2.py`

### Audio synthesis

- [Mel generation](https://github.com/iszhanjiawei/indextts2/tree/main/indextts/s2mel/modules): `indextts/s2mel/modules`
- [Codec components](https://github.com/iszhanjiawei/indextts2/tree/main/indextts/utils/maskgct/models/codec): `indextts/utils/maskgct/models/codec`
- [BigVGAN vocoder](https://github.com/iszhanjiawei/indextts2/blob/main/indextts/s2mel/modules/bigvgan/bigvgan.py): `indextts/s2mel/modules/bigvgan/bigvgan.py`

### Other components

- **TTS user**
- **Generated audio**

## Connections

- TTS user → Web interface: submits text (evidence: `README.md`)
- TTS user → Inference API: provides prompts (evidence: `README.md`)
- Web interface → Inference API: invokes inference (evidence: `webui.py`)
- Inference API → Text frontend: uses frontend (evidence: `indextts/infer_indextts2.py`)
- Inference API → Audio features: uses extractors (evidence: `indextts/infer_indextts2.py`)
- Inference API → Speaker encoding: uses encoder (evidence: `indextts/infer_indextts2.py`)
- Inference API → Emotion guidance: accepts guidance (evidence: `README.md`)
- Inference API → Autoregressive model: uses model (evidence: `indextts/infer_indextts2.py`)
- Autoregressive model → Conformer encoder: uses encoder (evidence: `indextts/gpt/model_v2.py`)
- Autoregressive model → Perceiver: uses perceiver (evidence: `indextts/gpt/model_v2.py`)
- Autoregressive model → Duration adaptation: controls duration (evidence: `README.md`)
- Inference API → Mel generation: uses synthesis modules (evidence: `indextts/infer_indextts2.py`)
- Inference API → Codec components: uses codec utilities (evidence: `indextts/infer_indextts2.py`)
- Inference API → BigVGAN vocoder: uses vocoder (evidence: `indextts/infer_indextts2.py`)
- Inference API → Generated audio: writes speech (evidence: `README.md`)
- BigVGAN vocoder → Generated audio: synthesizes waveform
