# CycleGAN-VC3: Examining and Improving CycleGAN-VCs for Mel-spectrogram Conversion

arXiv: https://arxiv.org/abs/2010.11672 · HTML: https://mlsanity.com/paper/2010.11672/

## Key facts

- Cited by 25 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 22 Oct 2020.

## Built on by (most cited first, 25 of 25)

- [StarGANv2-VC: A Diverse, Unsupervised, Non-parallel Framework for Natural-Sounding Voice Conversion](https://mlsanity.com/paper/2107.10394/) (2021)
- [iSTFTNet: Fast and Lightweight Mel-Spectrogram Vocoder Incorporating Inverse Short-Time Fourier Transform](https://mlsanity.com/paper/2203.02395/) (2022)
- [MaskCycleGAN-VC: Learning Non-parallel Voice Conversion with Filling in Frames](https://mlsanity.com/paper/2102.12841/) (2021)
- [Speech Representation Disentanglement with Adversarial Mutual Information Learning for One-shot Voice Conversion](https://mlsanity.com/paper/2208.08757/) (2022)
- [HiFTNet: A Fast High-Quality Neural Vocoder with Harmonic-plus-Noise Filter and Inverse Short Time Fourier Transform](https://mlsanity.com/paper/2309.09493/) (2023)
- [Deepfake CAPTCHA: A Method for Preventing Fake Calls](https://mlsanity.com/paper/2301.03064/) (2023)
- [Towards General-Purpose Text-Instruction-Guided Voice Conversion](https://mlsanity.com/paper/2309.14324/) (2023)
- [Pathological voice adaptation with autoencoder-based voice conversion](https://mlsanity.com/paper/2106.08427/) (2021)
- [RefineGAN: Universally Generating Waveform Better than Ground Truth with Highly Accurate Pitch and Intensity Responses](https://mlsanity.com/paper/2111.00962/) (2021)
- [Supervising 3D Talking Head Avatars with Analysis-by-Audio-Synthesis](https://mlsanity.com/paper/2504.13386/) (2025)
- [Wave-U-Net Discriminator: Fast and Lightweight Discriminator for Generative Adversarial Network-Based Speech Synthesis](https://mlsanity.com/paper/2303.13909/) (2023)
- [An Adaptive Learning based Generative Adversarial Network for One-To-One Voice Conversion](https://mlsanity.com/paper/2104.12159/) (2021)
- [Audio-based Kinship Verification Using Age Domain Conversion](https://mlsanity.com/paper/2410.11120/) (2024)
- [Axial Residual Networks for CycleGAN-based Voice Conversion](https://mlsanity.com/paper/2102.08075/) (2021)
- [Generative Adversarial Network based Voice Conversion: Techniques, Challenges, and Recent Advancements](https://mlsanity.com/paper/2504.19197/) (2025)
- [LatentVoiceGrad: Nonparallel Voice Conversion with Latent Diffusion/Flow-Matching Models](https://mlsanity.com/paper/2509.08379/) (2025)
- [Music Sentiment Transfer](https://mlsanity.com/paper/2110.05765/) (2021)
- [CAESynth: Real-Time Timbre Interpolation and Pitch Control with Conditional Autoencoders](https://mlsanity.com/paper/2111.05174/) (2021)
- [Enhancing Zero-Shot Many to Many Voice Conversion with Self-Attention VAE](https://mlsanity.com/paper/2203.16037/) (2022)
- [Boosting Star-GANs for Voice Conversion with Contrastive Discriminator](https://mlsanity.com/paper/2209.10088/) (2022)
- [MetaSpeech: Speech Effects Switch Along with Environment for Metaverse](https://mlsanity.com/paper/2210.13811/) (2022)
- [Voice Conversion with Denoising Diffusion Probabilistic GAN Models](https://mlsanity.com/paper/2308.14319/) (2023)
- [Parallel and Limited Data Voice Conversion Using Stochastic Variational Deep Kernel Learning](https://mlsanity.com/paper/2309.04420/) (2023)
- [MaskCycleGAN-based Whisper to Normal Speech Conversion](https://mlsanity.com/paper/2408.14797/) (2024)
- [Pureformer-VC: Non-parallel One-Shot Voice Conversion with Pure Transformer Blocks and Triplet Discriminative Training](https://mlsanity.com/paper/2409.01668/) (2024)

## Abstract (arXiv)

Non-parallel voice conversion (VC) is a technique for learning mappings between source and target speeches without using a parallel corpus. Recently, cycle-consistent adversarial network (CycleGAN)-VC and CycleGAN-VC2 have shown promising results regarding this problem and have been widely used as benchmark methods. However, owing to the ambiguity of the effectiveness of CycleGAN-VC/VC2 for mel-spectrogram conversion, they are typically used for mel-cepstrum conversion even when comparative methods employ mel-spectrogram as a conversion target. To address this, we examined the applicability of CycleGAN-VC/VC2 to mel-spectrogram conversion. Through initial experiments, we discovered that their direct applications compromised the time-frequency structure that should be preserved during conversion. To remedy this, we propose CycleGAN-VC3, an improvement of CycleGAN-VC2 that incorporates time-frequency adaptive normalization (TFAN). Using TFAN, we can adjust the scale and bias of the converted features while reflecting the time-frequency structure of the source mel-spectrogram. We evaluated CycleGAN-VC3 on inter-gender and intra-gender non-parallel VC. A subjective evaluation of naturalness and similarity showed that for every VC pair, CycleGAN-VC3 outperforms or is competitive with the two types of CycleGAN-VC2, one of which was applied to mel-cepstrum and the other to mel-spectrogram. Audio samples are available at http://www.kecl.ntt.co.jp/people/kaneko.takuhiro/projects/cyclegan-vc3/index.html.