Research map / Image generation / Autoregressive and latent generation
SongEcho: Towards Cover Song Generation via Instance-Adaptive Element-wise Linear Modulation
Sifei Li, Yang Li, Zizhou Wang, Yuxin Zhang, Fuzhang Wu, Oliver Deussen, Tong-Yee Lee, Weiming Dong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Replace element-wise addition with EiLM (ablation) | SongEcho · Suno70k test | RPA: 0.6336 -> 0.6799 | Improved |
| Add IACR on top of EiLM (ablation) | SongEcho · Suno70k test | FD openl3: 75.28 -> 42.06 (RPA 0.6799 -> 0.7080) | Improved |
| Insert IA-EiLM before FFN instead of self-attention (ablation) | SongEcho · Suno70k test | RPA: 0.6190 -> 0.7080 | Improved |
| Increase training samples from 100 to 1000 (ablation) | SongEcho · Suno70k test | RPA: 0.4677 -> 0.6505 | Improved |
| Use full training set instead of 1000 samples (ablation) | SongEcho · Suno70k test | PER: 0.2871 -> 0.2951 | No clear effect |
| Use SongEcho instead of ACE-Step+MuseControlLite | ACE-Step · Suno70k test | RPA: 0.5205 -> 0.7080 | Improved |
| Swap text tags at test time | SongEcho · Suno70k test | CLAP: 0.3243 -> 0.2674 | Hurt |
| Add SongEcho melody control on SongEval transcribed lyrics | ACE-Step · SongEval (94 songs) | PER: 0.4510 -> 0.4487 | No clear effect |
| Use MuseControlLite with lower melody guidance (7.0) instead of 15.0 (ablation) | ACE-Step+MuseControlLite · Suno70k test | RPA / FD: RPA 0.5205 -> 0.4896; FD 72.04 -> 55.82 | Mixed |
Closest papers by meaning
- SongGen: A Single Stage Auto-regressive Transformer for Text-to-Song Generation (ICML 2025 · similarity 0.90)
- LeVo: High-Quality Song Generation with Multi-Preference Alignment (NeurIPS 2025 · similarity 0.90)
- SongEditor: Adapting Zero-Shot Song Generation Language Model as a Multi-Task Editor (AAAI 2025 · similarity 0.88)
- MuseControlLite: Multifunctional Music Generation with Lightweight Conditioners (ICML 2025 · similarity 0.88)
- MusicRL: Aligning Music Generation to Human Preferences (ICML 2024 · similarity 0.87)
- Whole-Song Hierarchical Generation of Symbolic Music Using Cascaded Diffusion Models (ICLR 2024 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
