### Install System Dependencies

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs essential system dependencies for audio processing using apt.

```bash
apt install portaudio19-dev libsox-dev ffmpeg
```

--------------------------------

### Start Server with Docker Compose

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

This command starts the API server profile using Docker Compose. Ensure Docker and Docker Compose are installed and configured.

```bash
docker compose --profile server up
```

--------------------------------

### Start WebUI with Docker Compose (CUDA)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Use this command to start the WebUI with CUDA support. Ensure you have cloned the repository and are in the project directory.

```bash
git clone https://github.com/fishaudio/fish-speech.git
cd fish-speech

docker compose --profile webui up
```

--------------------------------

### Start API Server with Docker Compose (CUDA)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Use this command to start the API server with CUDA support. Ensure you have cloned the repository and are in the project directory.

```bash
docker compose --profile server up
```

--------------------------------

### Example .env file for Docker Compose

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

This example shows how to configure environment variables for Docker Compose deployments, including backend selection, compile optimization, and port settings.

```bash
# .env file example
BACKEND=cuda              # or cpu
COMPILE=1                 # Enable compile optimization
GRADIO_PORT=7860         # WebUI port
API_PORT=8080            # API server port
UV_VERSION=0.8.15        # UV package manager version
CUDA_VER=12.9.0          # CUDA base image version (e.g. 12.6.0 for older drivers)
UV_EXTRA=cu129           # PyTorch CUDA variant (cu126, cu128, cu129) — must match CUDA_VER
```

--------------------------------

### Install PyTorch with Intel XPU Support

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs a nightly build of PyTorch with Intel XPU support for GPU acceleration.

```bash
pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/xpu
```

--------------------------------

### Install and Configure Fish-Audio S2

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/index.md

Instructions for setting up the Fish-Audio S2 system for use with LLM agents. Refer to the provided URL for detailed installation steps.

```text
Install and configure Fish-Audio S2 by following the instructions here: https://speech.fish.audio/install/
```

--------------------------------

### Start Awesome WebUI Backend Server

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Starts the API server for the Awesome WebUI. The `--compile` flag enables CUDA kernel fusion for faster inference. The server listens on `0.0.0.0:8888`.

```bash
python tools/api_server.py --listen 0.0.0.0:8888 --compile
```

--------------------------------

### Build Awesome WebUI Dependencies

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Installs the necessary Node.js dependencies for the Awesome WebUI. Ensure Node.js and npm are installed.

```bash
cd awesome_webui
npm install
```

--------------------------------

### Dataset Structure Example

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Organize your dataset with audio files and corresponding .lab annotation files under speaker-specific directories.

```bash
.
├── SPK1
│   ├── 21.15-26.44.lab
│   ├── 21.15-26.44.mp3
│   ├── 27.51-29.98.lab
│   ├── 27.51-29.98.mp3
│   ├── 30.1-32.71.lab
│   └── 30.1-32.71.mp3
└── SPK2
    ├── 38.79-40.85.lab
    └── 38.79-40.85.mp3
```

--------------------------------

### Start API Server with Docker Compose (CUDA + Compile Optimization)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Starts the API server with CUDA and enables compile optimization. Ensure the repository is cloned and you are in the project directory.

```bash
COMPILE=1 docker compose --profile server up
```

--------------------------------

### Start WebUI with Docker Compose (CPU-only)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Starts the WebUI using only the CPU. This is suitable for deployments without a compatible GPU. Ensure the repository is cloned and you are in the project directory.

```bash
BACKEND=cpu docker compose --profile webui up
```

--------------------------------

### Troubleshoot PyAudio Installation with Conda

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs PyAudio using Conda if pip installation fails, then retries the Fish Speech installation.

```bash
# conda install pyaudio
# Then run pip install -e . again
```

--------------------------------

### Start WebUI with Docker Compose (CUDA + Compile Optimization)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Starts the WebUI with CUDA and enables compile optimization for potentially faster performance. Ensure the repository is cloned and you are in the project directory.

```bash
COMPILE=1 docker compose --profile webui up
```

--------------------------------

### Install Fish Speech with Conda (CPU)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs Fish Speech with CPU-only support using pip.

```bash
pip install -e .[cpu]
```

--------------------------------

### Install Fish Speech with Conda (Default)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs Fish Speech using the default PyTorch index.

```bash
pip install -e .
```

--------------------------------

### Start Local API Server

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Use this command to start the Fish Speech API server locally. Specify paths to your Llama and decoder checkpoints and the desired listen address.

```bash
python tools/api_server.py \
  --llama-checkpoint-path checkpoints/s2-pro \
  --decoder-checkpoint-path checkpoints/s2-pro/codec.pth \
  --listen 0.0.0.0:8080
```

--------------------------------

### Start WebUI with Docker Compose (ROCm)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Starts the WebUI using Docker Compose with ROCm support for AMD GPUs. This requires ROCm drivers and Docker with GPU passthrough configured.

```bash
docker compose -f compose.rocm.yml --profile webui up --build
```

--------------------------------

### Install C++ Standard Library for Intel Arc XPU

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs the required C++ standard library from conda-forge for Intel Arc XPU support.

```bash
conda install libstdcxx -c conda-forge
```

--------------------------------

### Install Fish Speech with UV (CPU)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Synchronizes dependencies using UV for Fish Speech with CPU-only support.

```bash
uv sync --python 3.12 --extra cpu
```

--------------------------------

### Start API Server with Docker Compose (ROCm)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Starts the API server using Docker Compose with ROCm support for AMD GPUs. This requires ROCm drivers and Docker with GPU passthrough configured.

```bash
docker compose -f compose.rocm.yml --profile server up --build
```

--------------------------------

### Conda Environment Setup

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Creates and activates a Conda environment for Fish Speech with Python 3.12.

```bash
conda create -n fish-speech python=3.12
conda activate fish-speech
```

--------------------------------

### Install Fish Speech with Conda (GPU)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Installs Fish Speech with GPU support using pip, specifying a CUDA version (e.g., cu129).

```bash
pip install -e .[cu129]
```

--------------------------------

### Install Fish Speech with UV (GPU)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Synchronizes dependencies using UV for Fish Speech with GPU support, specifying a CUDA version (e.g., cu129).

```bash
uv sync --python 3.12 --extra cu129
```

--------------------------------

### Download Model Weights

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Use this command to download the required model weights for the Fish Audio S2 model. Ensure you have `hf` installed.

```bash
hf download fishaudio/s2-pro --local-dir checkpoints/s2-pro
```

--------------------------------

### Python Client for Text-to-Speech

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Example of using the Python client to perform text-to-speech generation via the API server. This command uses the default model loaded on the server.

```bash
python tools/api_client.py \
  --url http://127.0.0.1:8080/v1/tts \
  --text "Hello from Fish Speech" \
  --output s2-pro-demo
```

--------------------------------

### Get VQ Tokens from Reference Audio

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

This script extracts VQ tokens from a reference audio file, which are necessary for subsequent text-to-semantic token generation. It outputs `fake.npy` and `fake.wav`.

```python
python fish_speech/models/dac/inference.py \
    -i "test.wav" \
    --checkpoint-path "checkpoints/s2-pro/codec.pth"
```

--------------------------------

### Pack Dataset into Protobuf

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Pack your prepared dataset into protobuf format using the build_dataset.py script. This is a necessary step before fine-tuning.

```bash
python tools/llama/build_dataset.py \
    --input "data" \
    --output "data/protos" \
    --text-extension .lab \
    --num-workers 16
```

--------------------------------

### Build Awesome WebUI

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Compiles the frontend assets for the Awesome WebUI.

```bash
npm run build
```

--------------------------------

### Run Gradio WebUI

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Launches the Gradio-based Web User Interface for inference. The `--compile` flag can be added for acceleration.

```bash
python tools/run_webui.py # --compile if you need acceleration
```

--------------------------------

### Download VQGAN Weights

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Download the necessary VQGAN weights before extracting semantic tokens. Ensure the path is correctly specified.

```bash
huggingface-cli download fishaudio/openaudio-s1-mini --local-dir checkpoints/openaudio-s1-mini
```

--------------------------------

### Fine-tuning LLAMA with LoRA

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Initiate the fine-tuning process for the LLAMA model using LoRA. Ensure LLAMA weights are downloaded and specify the LoRA configuration.

```bash
python fish_speech/train.py --config-name text2semantic_finetune \
    project=$project \
    +lora@model.model.lora_config=r_8_alpha_16
```

--------------------------------

### Run WebUI Inference

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

Launch the WebUI for inference. Use the '--compile' flag to fuse CUDA kernels for significantly faster inference.

```python
!python tools/run_webui.py \
    --llama-checkpoint-path checkpoints/openaudio-s1-mini \
    --decoder-checkpoint-path checkpoints/openaudio-s1-mini/codec.pth \
    # --compile
```

--------------------------------

### Build CPU-only Docker Images (Multi-platform)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Builds CPU-only Docker images for both WebUI and API server, supporting multiple platforms like amd64 and arm64. This is for deployments without GPU acceleration.

```bash
docker build \
    --platform linux/amd64,linux/arm64 \
    -f docker/Dockerfile \
    --build-arg BACKEND=cpu \
    --target webui \
    -t fish-speech-webui:cpu .
```

--------------------------------

### Run WebUI Container (ROCm)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Runs a manually built WebUI container with ROCm support. This command includes necessary device mappings and environment variables for AMD GPU acceleration.

```bash
docker run \
    --device=/dev/kfd --device=/dev/dri \
    --group-add video --group-add render \
    -e ROCBLAS_USE_HIPBLASLT=0 \
    -v ./checkpoints:/app/checkpoints \
    -p 7860:7860 \
    fish-speech-webui:rocm
```

--------------------------------

### Build WebUI Docker Image (CUDA)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Manually builds the Docker image for the WebUI with CUDA support. This command is for advanced users who need to customize the build process.

```bash
docker build \
    --platform linux/amd64 \
    -f docker/Dockerfile \
    --build-arg BACKEND=cuda \
    --build-arg CUDA_VER=12.9.0 \
    --build-arg UV_EXTRA=cu129 \
    --target webui \
    -t fish-speech-webui:cuda .
```

--------------------------------

### Build API Server Docker Image (CUDA)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Manually builds the Docker image for the API server with CUDA support. This command is for advanced users who need to customize the build process.

```bash
docker build \
    --platform linux/amd64 \
    -f docker/Dockerfile \
    --build-arg BACKEND=cuda \
    --build-arg CUDA_VER=12.9.0 \
    --build-arg UV_EXTRA=cu129 \
    --target server \
    -t fish-speech-server:cuda .
```

--------------------------------

### Python Client with Reference Voice

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Use this command to perform text-to-speech generation with a specific reference voice ID. This selects the voice, not the base TTS model.

```bash
python tools/api_client.py \
  --url http://127.0.0.1:8080/v1/tts \
  --text "Hello from Fish Speech" \
  --reference_id my-speaker \
  --output s2-pro-demo
```

--------------------------------

### Download Fish-Speech Model

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

Download the specified Fish-Speech model using the 'hf download' command. Consider using a mirror for faster downloads if needed.

```python
# For Chinese users, you probably want to use mirror to accelerate downloading
# !set HF_ENDPOINT=https://hf-mirror.com
# !export HF_ENDPOINT=https://hf-mirror.com 

!hf download fishaudio/openaudio-s1-mini --local-dir checkpoints/openaudio-s1-mini/
```

--------------------------------

### Encode Reference Audio to Prompt

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

Encode a reference audio file into a prompt file (fake.npy). This step is necessary for generating speech from text later. The output audio is saved as fake.wav.

```python
## Enter the path to the audio file here
src_audio = r"D:\PythonProject\vo_hutao_draw_appear.wav"

!python fish_speech/models/dac/inference.py \
    -i {src_audio} \
    --checkpoint-path "checkpoints/openaudio-s1-mini/codec.pth"
```

```python
from IPython.display import Audio, display
audio = Audio(filename="fake.wav")
display(audio)
```

--------------------------------

### Build Development Docker Image (CUDA)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Builds a development-focused Docker image with CUDA support. This image is suitable for local development and debugging purposes.

```bash
docker build \
    --platform linux/amd64 \
    -f docker/Dockerfile \
    --build-arg BACKEND=cuda \
    --target dev \
    -t fish-speech-dev:cuda .
```

--------------------------------

### Loudness Normalization

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Apply loudness normalization to your dataset using the fish-audio-preprocess tool. This step is recommended for better audio quality.

```bash
fap loudness-norm data-raw data --clean
```

--------------------------------

### Build WebUI Docker Image (ROCm)

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/install.md

Manually builds the Docker image for the WebUI with ROCm support for AMD GPUs. This command is used for advanced customization of the ROCm environment.

```bash
docker build -f docker/Dockerfile.rocm --target webui -t fish-speech-webui:rocm .
```

--------------------------------

### Generate Semantic Tokens from Text

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

Converts input text into semantic tokens using a reference text and its corresponding VQ tokens. The `--compile` flag can be used for faster inference, and `--half` is for GPUs that do not support bf16.

```python
python fish_speech/models/text2semantic/inference.py \
    --text "The text you want to convert" \
    --prompt-text "Your reference text" \
    --prompt-tokens "fake.npy" \
    # --compile
```

--------------------------------

### Generate Vocals from Semantic Tokens

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/inference.md

This script generates the final vocal audio file (`fake.wav`) from the previously generated semantic tokens.

```python
python fish_speech/models/dac/inference.py \
    -i "codes_0.npy" \

```

--------------------------------

### Enable React-Specific ESLint Rules

Source: https://github.com/fishaudio/fish-speech/blob/main/awesome_webui/README.md

Integrate 'eslint-plugin-react-x' and 'eslint-plugin-react-dom' for React and React DOM specific linting. Project configuration is similar to type-aware rules.

```javascript
// eslint.config.js
import reactX from 'eslint-plugin-react-x'
import reactDom from 'eslint-plugin-react-dom'

export default defineConfig([
  globalIgnores(['dist']),
  {
    files: ['**/*.{ts,tsx}'],
    extends: [
      // Other configs...
      // Enable lint rules for React
      reactX.configs['recommended-typescript'],
      // Enable lint rules for React DOM
      reactDom.configs.recommended,
    ],
    languageOptions: {
      parserOptions: {
        project: ['./tsconfig.node.json', './tsconfig.app.json'],
        tsconfigRootDir: import.meta.dirname,
      },
      // other options...
    },
  },
])
```

--------------------------------

### Generate Semantic Tokens from Text

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

Generate semantic tokens from input text, using a reference audio prompt. This command creates a 'codes_N.npy' file. Using '--compile' can increase inference speed.

```python
!python fish_speech/models/text2semantic/inference.py \
    --text "hello world" \
    --prompt-text "The text corresponding to reference audio" \
    --prompt-tokens "fake.npy" \
    --checkpoint-path "checkpoints/openaudio-s1-mini" \
    --num-samples 2
    # --compile
```

--------------------------------

### Batch Extraction of Semantic Tokens

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

Extract semantic tokens from your dataset using the provided Python script. Adjust `--num-workers` and `--batch-size` based on your GPU memory.

```bash
python tools/vqgan/extract_vq.py data \
    --num-workers 1 --batch-size 16 \
    --config-name "modded_dac_vq" \
    --checkpoint-path "checkpoints/openaudio-s1-mini/codec.pth"
```

--------------------------------

### Dataset Structure After Token Extraction

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

After token extraction, .npy files containing semantic tokens will be added to your data directory.

```bash
.
├── SPK1
│   ├── 21.15-26.44.lab
│   ├── 21.15-26.44.mp3
│   ├── 21.15-26.44.npy
│   ├── 27.51-29.98.lab
│   ├── 27.51-29.98.mp3
│   ├── 27.51-29.98.npy
│   ├── 30.1-32.71.lab
│   ├── 30.1-32.71.mp3
│   └── 30.1-32.71.npy
└── SPK2
    ├── 38.79-40.85.lab
    ├── 38.79-40.85.mp3
    └── 38.79-40.85.npy
```

--------------------------------

### Enable Type-Aware ESLint Rules

Source: https://github.com/fishaudio/fish-speech/blob/main/awesome_webui/README.md

Configure ESLint to use type-aware lint rules for TypeScript files. Ensure 'tsconfig.node.json' and 'tsconfig.app.json' are correctly specified.

```javascript
export default defineConfig([
  globalIgnores(['dist']),
  {
    files: ['**/*.{ts,tsx}'],
    extends: [
      // Other configs...

      // Remove tseslint.configs.recommended and replace with this
      tseslint.configs.recommendedTypeChecked,
      // Alternatively, use this for stricter rules
      tseslint.configs.strictTypeChecked,
      // Optionally, add this for stylistic rules
      tseslint.configs.stylisticTypeChecked,

      // Other configs...
    ],
    languageOptions: {
      parserOptions: {
        project: ['./tsconfig.node.json', './tsconfig.app.json'],
        tsconfigRootDir: import.meta.dirname,
      },
      // other options...
    },
  },
])
```

--------------------------------

### Generate Speech from Semantic Tokens

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

Synthesize speech from the generated semantic tokens (codes_N.npy). The output speech is saved as fake.wav.

```python
!python fish_speech/models/dac/inference.py \
    -i "codes_0.npy" \
    --checkpoint-path "checkpoints/openaudio-s1-mini/codec.pth"
```

```python
from IPython.display import Audio, display
audio = Audio(filename="fake.wav")
display(audio)
```

--------------------------------

### Set Linux Locale for UTF-8

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

For Linux users, set the locale to 'en_US.UTF-8' to handle character encoding correctly.

```python
import locale
locale.setlocale(locale.LC_ALL, 'en_US.UTF-8')
```

--------------------------------

### API Server Health Check

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Perform a health check on the running API server to ensure it is operational. The expected response indicates the server is ready.

```bash
curl -X GET http://127.0.0.1:8080/v1/health
```

--------------------------------

### Set Windows Console Code Page

Source: https://github.com/fishaudio/fish-speech/blob/main/inference.ipynb

For Windows users, set the console code page to UTF-8 to ensure proper character display.

```python
!chcp 65001
```

--------------------------------

### Merge LoRA Weights

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/finetune.md

After fine-tuning, merge the LoRA weights with the base model weights to create a consolidated model for inference. Specify the correct paths for base and LoRA weights.

```bash
python tools/llama/merge_lora.py \
	--lora-config r_8_alpha_16 \
	--base-weight checkpoints/openaudio-s1-mini \
	--lora-weight results/$project/checkpoints/step_000000010.ckpt \
	--output checkpoints/openaudio-s1-mini-yth-lora/
```

--------------------------------

### VQ Encode

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Encodes input data into a VQ representation.

```APIDOC
## VQ Encode

### Description
Encodes input data into a VQ representation.

### Method
POST

### Endpoint
/v1/vqgan/encode
```

--------------------------------

### Text-to-Speech Generation

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Generates speech from input text using the configured TTS model.

```APIDOC
## Text-to-Speech Generation

### Description
Generates speech from input text using the configured TTS model.

### Method
POST

### Endpoint
/v1/tts

### Request Body
- **text** (string) - Required - The text to convert to speech.
- **reference_id** (string) - Optional - The ID of a saved voice reference to use for style transfer.
```

--------------------------------

### VQ Decode

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Decodes a VQ representation back into its original data format.

```APIDOC
## VQ Decode

### Description
Decodes a VQ representation back into its original data format.

### Method
POST

### Endpoint
/v1/vqgan/decode
```

--------------------------------

### Health Check

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

Checks the status of the running API server.

```APIDOC
## Health Check

### Description
Checks the status of the running API server.

### Method
GET

### Endpoint
/v1/health

### Response
#### Success Response (200)
- **status** (string) - Indicates the server status, typically "ok".
```

--------------------------------

### API Server Health Check Response

Source: https://github.com/fishaudio/fish-speech/blob/main/docs/en/server.md

The expected JSON response for a successful health check of the API server.

```json
{"status":"ok"}
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.