Dialogue-Aware Video-to-Music Generation Using Public Domain Film Collections
Open Screen Soundtrack Library 2 (OSSL-v2)
A self-hosted corpus of 34,343 video clips totaling 246.4 hours, sourced from public-domain films. Every clip is paired with its source-separated music track.
| Video | Music | |
|---|---|---|
| Demo #1 | ||
| Demo #2 | ||
| Demo #3 |
Dialogue Adapter Demo
🔊 Please turn your volume down before playing. The Diff-V2M samples are loud and noisy. We recommend listening at a very low volume, and through speakers rather than earphones. 🎧🚫
Each clip is played with its original dialogue and the music generated for it. Only the music track differs between cells, and every generated track is loudness-matched to the clip's real score.
Example 1
| GVMGen | VidMuse | Diff-V2M | |
|---|---|---|---|
| Baseline | |||
| (+Dialogue) |
Example 2
| GVMGen | VidMuse | Diff-V2M | |
|---|---|---|---|
| Baseline | |||
| (+Dialogue) |