End-to-end speech recognition large model: 31 languages, dialects, accents, lyrics, hotwords, timestamps, speaker diarization. Trained on tens of millions of hours.

Open-source industrial-grade ASR models supporting Mandarin, Chinese dialects and English, achieving a new SOTA on public Mandarin ASR benchmarks, while also offering outstanding singing lyrics recognition capability.

Extract public Spotify data — tracks, albums, artists, playlists, podcasts & lyrics — without the official API. Sync + async, typed models, one dependency.

[CHI 2023 Honorable Mention] 音楽に合わせてタイミングよく歌詞が動くWebアプリケーション（リリックアプリ）を開発できるJavaScript API / JavaScript API for developing web applications that show lyrics synchronized with the music playback (lyric apps)

A multi-agent system designed for generating music videos with scrolling subtitles based on lyrics. This project analyzes the lyrics and creates detailed prompts based on the analysis results to generate story-like images, ultimately producing an image-to-image music video, using openai API, gpt-4o and dall-e 3 models.