Research map / Image generation / Autoregressive and latent generation
SING: Symbol-to-Instrument Neural Generator
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| use spectral loss instead of waveform loss (ablation) | SING · NSynth pitch completion test | spectral loss: 0.084 -> 0.051 | Helped |
| use spectral loss instead of waveform loss (ablation) | SING · NSynth pitch completion test | waveform MSE: 0.039 -> N/A | Mixed |
| remove the time embedding (ablation) | SING · NSynth pitch completion test | spectral loss: 0.051 -> 0.063 | Hurt |
| replace full SING with its convolutional autoencoder (ablation) | SING · NSynth pitch completion test | spectral loss: 0.032 -> 0.051 | Mixed |
| generate conditioned waveforms with SING rather than Wavenet | NSynth pitch completion test · NSynth pitch completion test | MOS: 2.85 ± 0.24 -> 3.55 ± 0.23 | Helped |
| generate conditioned waveforms with SING rather than Wavenet | NSynth pitch completion test · NSynth pitch completion test | ABX preference: 69.7% | Helped |
| use frame-by-frame SING generation rather than autoregressive Wavenet generation | NSynth pitch completion test · NSynth pitch completion test | generation speed: 0.2 sec/sec -> 512 sec/sec | Helped |
| use SING rather than Wavenet | NSynth pitch completion test · NSynth pitch completion test | training time (hrs * GPU): 3840 -> 120 | Helped |
Closest papers by meaning
- Neural Audio Synthesis of Musical Notes with WaveNet Autoencoders (ICML 2017 · similarity 0.91)
- Efficient Neural Music Generation (NeurIPS 2023 · similarity 0.89)
- PerformanceNet: Score-to-Audio Music Generation with Multi-Band Convolutional Residual Network (AAAI 2019 · similarity 0.89)
- Creative Text-to-Audio Generation via Synthesizer Programming (ICML 2024 · similarity 0.89)
- SongGen: A Single Stage Auto-regressive Transformer for Text-to-Song Generation (ICML 2025 · similarity 0.88)
- Deep Learning for Style Transfer and Experimentation with Audio Effects and Music Creation (AAAI 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.