ASR_test

A Python-based Automatic Speech Recognition (ASR) demo project that provides both a server API and a simple UI interface. It allows you to quickly deploy a local speech recognition service and test it through a front-end interface. Ideal for learning ASR API usage, deployment, and integration.

---

🚀 Features

• 🎙️ ASR API Server Provides a FastAPI-based REST interface for speech-to-text inference.

• 🧑‍💻 Cross-platform support Includes startup scripts for both Linux and Windows environments.

• 🖼️ Simple UI Demo Uses Gradio for a lightweight web interface, making testing and debugging easy.

• 📡 Integration-ready Can be connected to front-end applications or automated workflows to convert audio files to text.

---

📂 Project Structure

.
├── asr_api_linux.py        # ASR API server for Linux
├── asr_api_windows.py      # ASR API server for Windows
├── asr_ui.py               # Gradio web UI demo
└── README.md               # This document

---

🛠️ Environment & Dependencies

Recommended: Python 3.8+ with a virtual environment:

python3 -m venv asr_env
source asr_env/bin/activate     # Linux/macOS
asr_env\Scripts\activate        # Windows

Install dependencies:

pip install -U funasr fastapi uvicorn
pip install gradio requests

funasr is used for loading and running speech recognition models. You can replace it with other ASR frameworks if needed.

---

▶️ Running the ASR API Server

🐧 Linux

python -m uvicorn asr_api_linux:app --host 0.0.0.0 --port 8000

🪟 Windows

python -m uvicorn asr_api_windows:app --host 0.0.0.0 --port 8000

Once running, open your browser:

http://127.0.0.1:8000/docs

You can view the auto-generated API docs and test the endpoints.

---

🖥️ Running the UI Demo

After starting the API server, run:

python asr_ui.py

Open your browser at:

http://127.0.0.1:7860/

You can now test speech recognition through the web interface.

---

📌 API Usage Example

Assuming the API server is running:

curl -X POST "http://127.0.0.1:8000/asr" \
     -F "audio=@./sample.wav"

Sample response:

{
  "text": "This is the recognition result"
}

---

🧠 Custom Models

You can replace the ASR model according to your needs, including offline or faster models. If you change the model, make sure to update the funasr loading code in the scripts.

---

💡 Use Cases

• ✅ Local deployment of ASR service
• ✅ Speech-to-text integration
• ✅ ASR learning, experimentation, and demonstration
• ✅ Rapid prototyping of voice-enabled applications

---

❓ FAQ

Q: Can this be deployed on a cloud server? A: Yes, the API server can be deployed to the cloud, and the front-end can access it via the public IP.

Q: Does it support Chinese recognition? A: Yes, as long as you load a model that supports Chinese.