# DDSP-based Singing Vocoders: A New Subtractive-based Synthesizer and A Comprehensive Evaluation

arXiv: https://arxiv.org/abs/2208.04756 · HTML: https://mlsanity.com/paper/2208.04756/

## Key facts

- Cited by 11 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 9 Aug 2022.

## Built on by (most cited first, 11 of 11)

- [A Survey on Generative Diffusion Model](https://mlsanity.com/paper/2209.02646/) (2022)
- [Foundation Models for Music: A Survey](https://mlsanity.com/paper/2408.14340/) (2024)
- [The Ethical Implications of Generative Audio Models: A Systematic Literature Review](https://mlsanity.com/paper/2307.05527/) (2023)
- [GANStrument: Adversarial Instrument Sound Synthesis with Pitch-invariant Instance Conditioning](https://mlsanity.com/paper/2211.05385/) (2022)
- [HiddenSinger: High-Quality Singing Voice Synthesis via Neural Audio Codec and Latent Diffusion Models](https://mlsanity.com/paper/2306.06814/) (2023)
- [Unsupervised Harmonic Parameter Estimation Using Differentiable DSP and Spectral Optimal Transport](https://mlsanity.com/paper/2312.14507/) (2023)
- [Mandarin Singing Voice Synthesis with Denoising Diffusion Probabilistic Wasserstein GAN](https://mlsanity.com/paper/2209.10446/) (2022)
- [Anonymising Elderly and Pathological Speech: Voice Conversion Using DDSP and Query-by-Example](https://mlsanity.com/paper/2410.15500/) (2024)
- [InstrumentGen: Generating Sample-Based Musical Instruments From Text](https://mlsanity.com/paper/2311.04339/) (2023)
- [Generating Sample-Based Musical Instruments Using Neural Audio Codec Language Models](https://mlsanity.com/paper/2407.15641/) (2024)
- [Designing Neural Synthesizers for Low-Latency Interaction](https://mlsanity.com/paper/2503.11562/) (2025)

## Abstract (arXiv)

A vocoder is a conditional audio generation model that converts acoustic features such as mel-spectrograms into waveforms. Taking inspiration from Differentiable Digital Signal Processing (DDSP), we propose a new vocoder named SawSing for singing voices. SawSing synthesizes the harmonic part of singing voices by filtering a sawtooth source signal with a linear time-variant finite impulse response filter whose coefficients are estimated from the input mel-spectrogram by a neural network. As this approach enforces phase continuity, SawSing can generate singing voices without the phase-discontinuity glitch of many existing vocoders. Moreover, the source-filter assumption provides an inductive bias that allows SawSing to be trained on a small amount of data. Our experiments show that SawSing converges much faster and outperforms state-of-the-art generative adversarial network and diffusion-based vocoders in a resource-limited scenario with only 3 training recordings and a 3-hour training time.