New benchmark for automatic lyric transcription

AudioShake’s research team released a paper on formatting-aware lyric transcriptions with unprecedented accuracy.

The AudioShake Research team found that public ALT benchmarks have focused exclusively on word content while ignoring the finer nuances of written lyrics including punctuation, line breaks, letter case, and non-word vocal sounds. These elements — which are implemented in the guidelines of music industry leaders for lyrics including Apple, LyricFind, and Musixmatch — are important for high-quality lyric transcripts as they make lyrics more readable and help convey rhythm, emotional emphasis, rhyme, etc. To address these issues, our team has introduced Jam-ALT, a new lyrics transcription benchmark that implements our compiled annotation guide, which unifies industry guidelines for lyrics.

Read the full blog post on AudioShake →

Read the full paper on arXiv →

🔎 Project website: https://audioshake.github.io/jam-alt/

🤗 Dataset: https://huggingface.co/datasets/audioshake/jam-alt

🧑‍💻 Code: https://github.com/audioshake/alt-eval/

AudioShake regularly works with labels, producers, publishers, and more to help open up songs to new possibilities. Any individuals or organizations looking to create clean stems with our technology can contact us directly at info@audioshake.ai.