MVPrompt: Building Music-Visual Prompts for AI Artists to Craft Music Video Mise-en-scène

要旨

Music videos have traditionally been the domain of experts, but with text-to-video generative AI models, AI artists can now create them more easily. However, accurately reflecting the desired music-visual mise-en-scène remains challenging without specialized knowledge, highlighting the need for supportive tools. To address this, we conducted a design workshop with seven music video experts, identified design goals, and developed MVPrompt—a tool for generating music-visual mise-en-scène prompts. In a user study with 24 AI artists, MVPrompt outperformed the Baseline, effectively supporting the collaborative creative process. Specifically, the Visual Theme stage facilitated the exploration of tone and manner, while the Visual Scene & Grammar stage refined prompts with detailed mise-en-scène elements. By enabling AI artists to specify mise-en-scène creatively, MVPrompt enhances the experience of making music video scenes with text-to-video generative AI.

著者
ChungHa Lee
Gwangju Institute of Science and Technology, Gwangju, Korea, Republic of
DaeHo Lee
Gwangju Institute of Science and Technology, Gwangju, Korea, Republic of
Jin-Hyuk Hong
Gwangju Institute of Science and Technology, Gwangju, Korea, Republic of
DOI

10.1145/3706598.3713876

論文URL

https://dl.acm.org/doi/10.1145/3706598.3713876

動画

会議: CHI 2025

The ACM CHI Conference on Human Factors in Computing Systems (https://chi2025.acm.org/)

セッション: Music

G314+G315
7 件の発表
2025-04-28 23:10:00
2025-04-29 00:40:00
日本語まとめ
読み込み中…