40k Songs with Audio Features and Lyrics
Multimodal dataset combining song lyrics and extracted audio characteristics, covering 79 musical genres and approximately 43,000 titles, ideal for music analysis projects and multimodal learning.
Around 43,000 songs in English with lyrics files (text) and audio features (CSV/JSON)
Apache 2.0
Description
The dataset 40k Songs with Audio Features and Lyrics brings together approximately 43,000 songs in English, including lyrics and explicit audio characteristics from three sources combined. It covers 79 varied musical genres, allowing a rich exploration of multimodal musical data.
What is this dataset for?
- Musical analysis and classification based on lyrics and audio characteristics
- Training multimodal models combining audio and text
- Search by automatically generating music or lyrics
Can it be enriched or improved?
The dataset can be supplemented with additional annotations such as mood, tempo, or more accurate metadata. The integration of complete raw audio data would also broaden its use.
🔎 In summary
🧠 Recommended for
- AI music researchers
- Multimodal projects
- NLP + audio developers
🔧 Compatible tools
- Librosa
- Hugging Face Transformers
- PyTorch
- TensorFlow
- SpacY
💡 Tip
Carefully pretreating speech (cleaning and tokenization) optimizes the performance of multimodal models.
Frequently Asked Questions
What languages are present in this dataset?
Mostly English, with a wide range of musical genres.
Can we use this dataset to generate music?
Yes, especially for training multimodal models combining audio and text.
Does the Apache 2.0 license allow commercial use?
Yes, this license is permissive and allows free commercial use.
