New dataset OSSL-v2 enhances reproducible video-to-music generation

Researchers have introduced the Open Screen Soundtrack Library version 2 (OSSL-v2), a reproducible dataset of 34,343 video clips totaling 246.4 hours, derived from public-domain films. This new corpus aims to address the reproducibility gap in video-to-music generation models, which are often trained on ephemeral YouTube links. The team also developed a method to use dialogue as a conditioning signal for video-to-music generation, enhancing existing models by incorporating dialogue tracks to better align music with on-screen speech. Their approach demonstrated improvements over state-of-the-art baselines when evaluated on both public-domain and commercial films. AI

IMPACT This research introduces a more robust dataset for training AI models in video-to-music generation, potentially improving the quality and emotional resonance of AI-generated soundtracks.

RANK_REASON The item describes a new dataset and a novel method for video-to-music generation presented in an academic paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Gotit.pub
• Hugging Face
• Open Screen Soundtrack Library version 2
• OSSL-v2
• ScienceCast
• YouTube

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →