---
title: AudioLDM2 SFX Generator
emoji: 🔊
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
pinned: false
short_description: Generate sound effects from text using AudioLDM2
---

# 🔊 AudioLDM2 SFX Generator

Generate high-quality sound effects from text descriptions using the **AudioLDM2** model.

## Features
- Text-to-SFX generation (whoosh, impacts, risers, ambient, etc.)
- Adjustable duration (1-15 seconds)
- Quality control via inference steps and guidance scale
- Negative prompts to avoid unwanted sounds
- Auto-selects best candidate from multiple generations

## API Usage (Gradio Client)

```python
from gradio_client import Client

client = Client("gaandrade99/audioldm2-sfx-generator")
result = client.predict(
    prompt="cinematic whoosh",
    negative_prompt="Low quality, distorted.",
    duration=5,
    steps=100,
    guidance_scale=3.5,
    num_candidates=1,
    api_name="/generate_sfx"
)
print(result)  # Path to generated WAV file
```

## Model
Uses [cvssp/audioldm2](https://huggingface.co/cvssp/audioldm2) - a 1.1B parameter text-to-audio diffusion model.