---
language:
- en
pipeline_tag: text-to-audio
tags:
- text-to-audio
license: other
---
# Improving Text-To-Audio Models with Synthetic Captions

🎵 We propose an audio captioning pipeline that uses an audio language model to synthesize accurate and diverse captions for audio at scale. We leverage this pipeline to produce a dataset of synthetic captions for AudioSet, named AF-AudioSet. We then pre-train our Tango family of text-to-audio models on these synthetic captions. 🎶

**This checkpoint was fine-tuned on the MusicCaps dataset**

[Read the paper](https://arxiv.org/pdf/2406.15487)

## Code

Our code is released here: [https://github.com/declare-lab/tango](https://github.com/declare-lab/tango)


Please follow the instructions in the repository for installation, usage and experiments.

## Quickstart Guide

Download the model and generate music from a text prompt:

```python
import IPython
import soundfile as sf
from tango import Tango

tango = Tango("declare-lab/tango-music-af-ft-mc")

prompt = "The song has a traditional jazzy feel to it. The piano chord progression is bouncy and light. The electric guitar has a chorus applied to it, and we hear various licks being played."
audio = tango.generate(prompt)
sf.write(f"{prompt}.wav", audio, samplerate=16000)
IPython.display.Audio(data=audio, rate=16000)
```

The model will be automatically downloaded and saved in cache. Subsequent runs will load the model directly from cache.

The `generate` function uses 100 steps by default to sample from the latent diffusion model. We recommend using 200 steps for generating better quality audios. This comes at the cost of increased run-time.

```python
prompt = "The song has a traditional jazzy feel to it. The piano chord progression is bouncy and light. The electric guitar has a chorus applied to it, and we hear various licks being played."
audio = tango.generate(prompt, steps=200)
IPython.display.Audio(data=audio, rate=16000)
```


Use the `generate_for_batch` function to generate multiple audio samples for a batch of text prompts:

```python
prompts = [
    "This song is a fusion of alternative and folk genres, highlighting simple yet soulful melodies and minimalist instrumentals.",
    "The instrumental music features an ensemble that resembles the orchestra. The melody is being played by a brass section while strings provide harmonic accompaniment.",
    "This music is instrumental. The tempo is fast with a lively keyboard harmony, steady drumming, groovy bass lines and harmonica melodic. The song is fresh, groovy, sunny, happy; vivacious and spirited."
]
audios = tango.generate_for_batch(prompts, samples=2)
```
This will generate two samples for each of the three text prompts.

## Citation
Please consider citing the following article if you found our work useful:
```
@article{kong2024improving,
  title={Improving Text-To-Audio Models with Synthetic Captions},
  author={Kong, Zhifeng and Lee, Sang-gil and Ghosal, Deepanway and Majumder, Navonil and Mehrish, Ambuj and Valle, Rafael and Poria, Soujanya and Catanzaro, Bryan},
  journal={arXiv preprint arXiv:2406.15487},
  year={2024}
}
```