CycleGAN-VC: Non-parallel Voice Conversion Using Cycle-Consistent Adversarial Networks
The DOI matches its Crossref record; 60 of the 61 references it cites were found in a registry or an index.
Abstract
We propose a non-parallel voice-conversion (VC) method that can learn a mapping from source to target speech without relying on parallel data. The proposed method is particularly noteworthy in that it is general purpose and high quality and works without any extra data, modules, or alignment procedure. Our method, called CycleGAN-VC, uses a cycle-consistent adversarial network (CycleGAN) with gated convolutional neural networks (CNNs) and an identity-mapping loss. A CycleGAN learns forward and inverse mappings simultaneously using adversarial and cycle-consistency losses. This makes it possible to find an optimal pseudo pair from non-parallel data. Furthermore, the adversarial loss can bring the converted speech close to the target one on the basis of indistinguishability without explicit density estimation. This allows to avoid over-smoothing caused by statistical averaging, which occurs in many conventional statistical model-based VC methods that represent data distribution explicitly. We configure a CycleGAN with gated CNNs and train it with an identity-mapping loss. This allows the mapping function to capture sequential and hierarchical structures while preserving linguistic information. We evaluated our method on a non-parallel VC task. An objective evaluation showed that the converted feature sequence was near natural in terms of global variance and modulation spectra, which are structural indicators highly correlated with subjective evaluation. A subjective evaluation showed that the quality of the converted speech was comparable to that obtained with a Gaussian mixture model-based parallel VC method even though CycleGAN-VC is trained under disadvantageous conditions (non-parallel and half the amount of data).
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“Recent research [59,60,61,62,63] in VC technologies has addressed long-standing challenges such as data scarcity, cross-lingual conversion, and real-time processing.”
From Audio Deepfake Detection: What Has Been Achieved and What Lies Ahead (2025) · cites this paper as [59]
Citations
Cites
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.90 · Cited by 235,474 · cited 1×
“Inspired by a previous study [49] for neural style transfer and super-resolution, we used the network that included downsampling, residual [50], and upsampling layers, as well as incorporating instance normalization [51].”
MethodAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946 · cited 1×
“We trained the network using the Adam optimizer [53] with a batch size of 1.”
MethodFully convolutional networks for semantic segmentation
Long, J., Shelhamer, E., & Darrell, T. (2015) · 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2015.7298965 · Cited by 38,834
Unpaired image-to-image translation using cycle-consistent adversarial networks
Zhu, J.-Y., Park, T., Isola, P., & Efros, A. A. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.244 · Cited by 23,602
Labaca-Castro, R. (2023) · Machine Learning under Malware Attack · doi:10.1007/978-3-658-40442-0_9 · Cited by 19,272
GAN(Generative Adversarial Nets)
(2017) · Journal of Japan Society for Fuzzy Theory and Intelligent Informatics · doi:10.3156/jsoft.29.5_177_2 · Cited by 19,270
(2010) · International Conference on Machine Learning · Cited by 13,995
Perceptual Losses for Real-Time Style Transfer and Super-Resolution
Johnson, J., Alahi, A., & Fei-Fei, L. (2016) · Lecture Notes in Computer Science · doi:10.1007/978-3-319-46475-6_43 · Cited by 11,283 · cited 2×
“Inspired by a previous study [49] for neural style transfer and super-resolution, we used the network that included downsampling, residual [50], and upsampling layers, as well as incorporating instance normalization [51].”
MethodShi, W., Caballero, J., Huszar, F., Totz, J., Aitken, A. P., Bishop, R., Rueckert, D., & Wang, Z. (2016) · 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr.2016.207 · Cited by 7,570
“Least squares generative adversarial networks,”
Mao, X., Li, Q., Xie, H., Lau, R. Y. K., Wang, Z., & Smolley, S. P. (2017) · 2017 IEEE International Conference on Computer Vision (ICCV) · doi:10.1109/iccv.2017.304 · Cited by 5,437 · cited 1×
“To stabilize training, we used a least squares GAN [52], which replaces the negative log likelihood objective in Ladv by a least squares loss.”
Method
Research works printed but not linked: 1
- maas, 2013, Rectifier nonlinearities improve neural network acoustic models, Proc ICML Workshop
Cited by
Items 201–300 of 354
DR-TiST: Disentangled Representation for Time Series Translation Across Application Domains
Arnout, H., Bronner, J., Kehrer, J., & Runkler, T. A. (2020) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn48605.2020.9207722 · Cited by 4
Non-Parallel Voice Conversion with Fewer Labeled Data by Conditional Generative Adversarial Networks
Chen, M., Hou, W., Ma, J., Wang, S., & Xiao, J. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-2162 · Cited by 4
“Among them, [16, 17] uses CycleGAN to take the acoustic sequence as an image with one channel and successfully performs VC with GAN.”
MethodNon-parallel Voice Conversion Based on Perceptual Star Generative Adversarial Network
Li, Y., Qiu, X., Cao, P., Zhang, Y., & Bao, B. (2022) · Circuits Systems and Signal Processing · doi:10.1007/s00034-022-01998-5 · Cited by 3
Speech Enhancement-assisted Voice Conversion in Noisy Environments
Chan, Y.-J., Peng, C.-J., Wang, S., Wang, H., Tsao, Y., & Chi, T.-S. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9979851 · Cited by 5
“For example, generative adversarial network (GAN) based approaches, such as CycleGAN [10] and StarGAN [11], apply a generator composed of an encoder and a decoder to perform VC, while using an additional discriminator or classifier to handle the statistical distribution of the target speaker.”
MethodSeki, S., Kameoka, H., Kaneko, T., & Tanaka, K. (2023) · IEEE Access · doi:10.1109/access.2023.3270699 · Cited by 7
“consistent GAN (CycleGAN) [11], [12], [13], [14] and a many-to-many extension of CycleGAN called StarGAN [15],”
BackgroundSelf-citationAny-to-Any Voice Conversion With Multi-Layer Speaker Adaptation and Content Supervision
Xu, X., Shi, L., Chen, X., Lin, P., Lian, J., Chen, J., Zhang, Z., & Hancock, E. R. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3306716 · Cited by 5
“Theselimitationshavemotivatedresearchers toexplorenon-parallelvoiceconversionapproaches[7],[8],[9], [10],whichledtoadeepneuralnetworkthatapproximateda mappingfunctionfromthesourcespeakerdomaintothetarget speakerdomain.”
BackgroundMethodInfluentialComparison of Data Augmentation Techniques on Filipino ASR for Children’s Speech
Abion, C., Lumapag, N. C., Ramirez, J. C., Resulto, C., & Lucas, C. R. (2023) · International Conference on Speech Technology and Human-Computer Dialogue (SpeD) · doi:10.1109/sped59241.2023.10314952 · Cited by 3
“Lastly, Voice Conversion (VC) techniques where linguistic information is retained while modifying non/para-linguistic information of the speech [17] were also used in augmenting data.”
BackgroundChoosing only the best voice imitators: Top-K many-to-many voice conversion with StarGAN
Fernandez‐Martín, C., Colomer, A., Panariello, C., & Naranjo, V. (2023) · Speech Communication · doi:10.1016/j.specom.2023.103022 · Cited by 5
An Improved System for Partially Fake Audio Detection Using Pre-trained Model
Zhang, J., Liu, H., Deng, M., Wang, J., Sun, Y. G., Xu, L., & Li, J. (2024) · Communications in computer and information science · doi:10.1007/978-981-97-0601-3_30 · Cited by 3
Enhancing Zero-Shot Multi-Speaker TTS with Negated Speaker Representations
Jeon, Y., Kim, Y., & Lee, G. G. (2024) · Proceedings of the AAAI Conference on Artificial Intelligence · doi:10.1609/aaai.v38i16.29793 · Cited by 9
Mahmud, S., Chowdhury, M. E. H., Chowdhury, M. H., Alqahtani, A., Mahbub, Z. B., Bensaali, F., & Serkan Kiranyaz, B. (2024) · Engineering Applications of Artificial Intelligence · doi:10.1016/j.engappai.2024.108483 · Cited by 3
Research progress on speech deepfake and its detection techniques
Xu, Y., Li, B., Tan, S., & Huang, J. (2024) · Journal of Image and Graphics · doi:10.11834/jig.230476 · Cited by 4
MAIN-VC: Lightweight Speech Representation Disentanglement for One-shot Voice Conversion
Li, P., Wang, J., Zhang, X., Zhang, Y., Xiao, J. Z., & Cheng, N. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10649929 · Cited by 4
“Inspired by image style transfer in computer vision, a series of methods based on generative adversarial network (GAN) have been applied to voice conversion [4]–[7].”
MethodSelf-Supervised Audio-Visual Soundscape Stylization
Li, T., Wang, R., Huang, P.-Y., Owens, A., & Anumanchipalli, G. (2024) · Lecture notes in computer science · doi:10.1007/978-3-031-72989-8_2 · Cited by 8
Hsu, W.-S., Lin, G., & Wang, W.-H. (2024) · Diagnostics · doi:10.3390/diagnostics14232693 · Cited by 4
Non-parallel Many-to-many Singing Voice Conversion by Adversarial Learning
Hu, J., Yu, C., & Guan, F. (2019) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc47483.2019.9023357 · Cited by 3
“[14] T. Kaneko and H. Kameoka, "CycleGAN-VC: Nonparallel Voice Conversion Using Cycle-Consistent Adversarial Networks," 2018 26th European Signal Processing Conference, Rome, 2018, pp. 2100-2104.”
MethodResultInfluentialKishida, T., & Nakashika, T. (2020) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2020edp7032 · Cited by 3
“CycleGAN-VC [2], [3] and StarGAN-VC [4], [5] are successful VC models incorporating GAN variants.”
MethodSemi-Supervised Learning for Singing Synthesis Timbre
Bonada, J., & Blaauw, M. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9413400 · Cited by 6
Saeki, T., Saito, Y., Takamichi, S., & Saruwatari, H. (2021) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2020edp7252 · Cited by 3
“Our additional task would be extending our system to other VC frameworks, including non-parallel training [39], [40] and multi-speaker conversion with speaker adaptation techniques [41]. tive evaluations, we can infer the effect of GV compensation in our system is limited.”
MethodOn the Potential of Modular Voice Conversion for Virtual Agents
Mertes, S., Kiderle, T., Schlagowski, R., Lingenfelser, F., & André, E. (2021) · International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) · doi:10.1109/aciiw52867.2021.9666349 · Cited by 5
Taco-VC: A Single Speaker Tacotron based Voice Conversion with Limited DataPreprint
Leshem, R. L., & Giryes, R. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1904.03522 · Cited by 3
Tanaka, M., Nose, T., Kanagaki, A., Shimizu, R., & Ito, A. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2005.03334 · Cited by 5
“Recently the CycleGAN-based parallel-data-free VC technique using convolutional neural networks (CNNs) was proposed (CycleGAN-VC) [9].”
MethodResultInfluentialDu, Z., Şişman, B., Zhou, K., & Li, H. (2021) · arXiv (Cornell University) · Cited by 3
ConTuner: Singing Voice Beautifying with Pitch and Expressiveness Condition
Wang, J., Li, P., Zhang, X., Cheng, N., & Xiao, J. Z. (2024) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn60899.2024.10650210 · Cited by 4
“Expressiveness Enhancer MIDI Spectrum Denoiser”
Singing to speech conversion with generative flow
Huang, J., & Benetos, E. (2025) · EURASIP Journal on Audio Speech and Music Processing · doi:10.1186/s13636-025-00400-x · Cited by 4
LatentVoiceGrad: Nonparallel Voice Conversion With Latent Diffusion/Flow-Matching Models
Kameoka, H., Kaneko, T., Tanaka, K., & Kondo, Y. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3613926 · Cited by 4
“Anon-parallelVCmethodwasproposedusingCycle-GAN[32],[33],[34],[35],avariantofGANoriginallydesigned forunpairedimagetranslation[36],[37],[38].”
Self-citationKermanshahi, M. A., Akbari, A., Nasersharif, B., & Kabudian, S. J. (2025) · Computers & Electrical Engineering · doi:10.1016/j.compeleceng.2025.110851 · Cited by 5
Change Your Singer: A Transfer Learning Generative Adversarial Framework for Song to Song Conversion
Daher, R., Zein, M. K., Zini, J. E., Awad, M., & Asmar, D. (2020) · International Joint Conference on Neural Networks (IJCNN) · doi:10.1109/ijcnn48605.2020.9206878 · Cited by 2
“Recently, Kaneko and Kameoka in [9] proposed a novel voice converter (CycleGAN-VC) that relies on cycle-consistent adversarial networks (CycleGAN).”
BackgroundMethodInfluentialAudio Noise Filter using Cycle Consistent Adversarial Network - CycleGAN ANF
Nguyen, N. S., Li, T., Zhang, X., Sheng, B., Wang, T., & Wang, J. (2019) · IEEE International Conference on Computer and Communications (ICCC) · doi:10.1109/iccc47050.2019.9064433 · Cited by 2
“Another direction involves deep learning on voice converter using parallel or unparalleled data [5], [6], [7], [8] especially the solution of CycleGAN VC [9], which also adopts CycleGAN architecture like our own.”
BackgroundNonparallel Training of Exemplar-Based Voice Conversion System Using INCA-Based Alignment Technique
Suda, H., Kotani, G., & Saito, D. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-1232 · Cited by 2
An evaluation of voice conversion with neural network spectral mapping models and WaveNet vocoder
Tobing, P. L., Wu, Y.-C., Hayashi, T., Kobayashi, K., & Toda, T. (2020) · APSIPA Transactions on Signal and Information Processing · doi:10.1017/atsip.2020.24 · Cited by 2
“Note that different from recent works on the use of non-parallel (unpaired) data, such as with restricted Boltzmann machine [33], variational autoencoder [34], or generative adversarial network [35], in this work, we focus on the development of NN-based VC with parallel data.”
MethodSaito, Y., Nakamura, T., Ijima, Y., Nishida, K., & Takamichi, S. (2020) · Nippon Onkyo Gakkaishi/Acoustical science and technology/Nihon Onkyo Gakkaishi · doi:10.1250/ast.42.1 · Cited by 3
Dorai, Y., Gazzah, S., Chausse, F., & Amara, N. E. B. (2021) · Pattern Analysis and Applications · doi:10.1007/s10044-021-00990-0 · Cited by 2
Voice Conversion Using Learnable Similarity-Guided Masked Autoencoder
Gu, Y., Zhao, X., Yi, X., & Xiao, J. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-25115-3_4 · Cited by 3
Beyond Text-to-Speech Synthesis
Tan, X. (2023) · Artificial intelligence: foundations, theory, and algorithms/Artificial intelligence: Foundations, theory, and algorithms · doi:10.1007/978-981-99-0827-1_12 · Cited by 2
One-Shot Voice Conversion Based on Style Generative Adversarial Networks with ESR and DSNet
Li, Y., Pan, L., Qiu, X., Yang, Z., Tan, Z., & Qian, B. (2024) · Circuits Systems and Signal Processing · doi:10.1007/s00034-024-02675-5 · Cited by 2
End-to-End Neural Formant Synthesis Using Low-Dimensional Acoustic Parameters
Kobayashi, S., Kosaka, T., & Nose, T. (2024) · IEEE Global Conference on Consumer Electronics (GCCE) · doi:10.1109/gcce62371.2024.10760681 · Cited by 2
ARVC: An Auto-Regressive Voice Conversion System Without Parallel Training Data
Lian, Z., Wen, Z., Zhou, X., Pu, S., Zhang, S., & Tao, J. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-1715 · Cited by 2
“To address this problem, generative adversarial network (GAN) [13] and its variants (such as CycleGAN [14, 15] and StarGAN [16, 17]) use a discriminator that amplifies this artifact in the loss function.”
Simultaneous Conversion of Speaker Identity and Emotion Based on Multiple-Domain Adaptive RBM
Kishida, T., Shin, T., & Nakashika, T. (2020) · Interspeech (USB) · doi:10.21437/interspeech.2020-2262 · Cited by 2
“Recent research trends have focused on artificial neural network models such as a restricted Boltzmann machine (RBM) [12], a variational autoencoder (VAE) [13, 14], and a generative adversarial net (GAN) [15, 16, 17, 18].”
BackgroundVoice Privacy with Smart Digital Assistants in Educational Settings
Niknazar, M., Vempaty, A., & Kokku, R. (2021) · Lecture notes in computer science · doi:10.1007/978-3-030-80421-3_31 · Cited by 2
TVQVC: Transformer Based Vector Quantized Variational Autoencoder with CTC Loss for Voice Conversion
Chen, Z., & Zhang, P. (2021) · Interspeech (USB) · doi:10.21437/interspeech.2021-1301 · Cited by 4
“Generative adversarial network[9] and its variants, for example CycleGAN-VC[10, 11] and StarGAN-VC[12, 13], which use generator or conditional generator to transform the source speaker’s features to target speaker’s features directly.”
BackgroundHumming-to-Instrument Conversion based on CycleGAN
Lai, W.-H., Wang, S.-L., & Xu, Z.-Y. (2021) · International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS) · doi:10.1109/ispacs51563.2021.9651076 · Cited by 2
Translation of Time Series Data from Controlled DC Motors using Disentangled Representation Learning
Arnout, H., Bronner, J., Kehrer, J., & Runkler, T. A. (2021) · IEEE Symposium Series on Computational Intelligence (SSCI) · doi:10.1109/ssci50451.2021.9660007 · Cited by 2
Parallel data free singing voice conversion with cycle-consistent BEGAN
Yousuf, A., & George, D. S. (2022) · Materials Today Proceedings · doi:10.1016/j.matpr.2022.01.169 · Cited by 3
Arbitrary Voice Conversion via Adversarial Learning and Cycle Consistency Loss
Lian, J., Lin, P., Dai, Y., & Li, G. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-13829-4_49 · Cited by 2
Research on synthesis of designated speaker speech based on StarGAN-VC model
Qiu, X., & Luo, Y. (2022) · International Conference on Artificial Intelligence and Intelligent Information Processing (AIIIP) · doi:10.1117/12.2659719 · Cited by 2
Applying Generative Adversarial Networks and Vision Transformers in Speech Emotion Recognition
Heracleous, P., Fukayama, S., Ogata, J., & Mohammad, Y. F. O. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-17618-0_6 · Cited by 2
Short-Utterance-Based Children’s Speaker Verification in Low-Resource Conditions
Aziz, S., Ankita, A., & Shahnawazuddin, S. (2023) · Circuits Systems and Signal Processing · doi:10.1007/s00034-023-02535-8 · Cited by 4
You, C. H., & Dong, M. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10446264 · Cited by 2
“This characteristics of non-parallel VC makes itself suitable for cross-lingual VC. CycleGAN-VC is a typical non-parallel voice conversion (VC) method [10], it employs a sequential mapping approach through the utilization of cycle-consistent generative adversarial network (CycleGAN) [11] [12] [13].”
MethodHow Far Are We from Robust Voice Conversion: A SurveyPreprint
Huang, T.-H., Lin, J.-H., Huang, C., & Lee, H.-Y. (2020) · arXiv (Cornell University) · doi:10.48550/arxiv.2011.12063 · Cited by 2
“Recently, several models were proposed to utilize non-parallel data [1, 2, 3, 4, 5].”
BackgroundMPFM-VC: A Voice Conversion Algorithm Based on Multi-Dimensional Perception Flow Matching
Wang, Y., Han, X., Lv, S., Zhou, T., & Chu, Y. (2025) · Applied Sciences · doi:10.3390/app15105503 · Cited by 1
Robust analog/RF circuit design via Cycle-Consistent Generative Adversarial Networks
Guo, N., Tao, J., Zeng, X., & Li, X. (2025) · Integration · doi:10.1016/j.vlsi.2025.102488 · Cited by 1
Afiuny, N. S., & Lakizadeh, A. (2025) · EURASIP Journal on Audio Speech and Music Processing · doi:10.1186/s13636-025-00422-5 · Cited by 1
“Among these, CycleGAN [24] and its variants (e.g., CycleGAN-VC [25], CycleGAN-VC2 [26], CycleGAN-VC3 [27]) became prominent in non-parallel VC by leveraging cycle-consistency loss to learn mappings between unpaired domains [23].”
A Progressive Generation Framework with Speech Pre-trained Model for Expressive Voice Conversion
Wang, T., Ge, M., Niu, Z., Gong, C., Qiang, C., Wang, H., Huang, Z., Ma, Z., Wang, X., Chen, X., et al. (2025) · IEEE International Conference on Multimedia and Expo (ICME) · doi:10.1109/icme59968.2025.11209157 · Cited by 1
“The DFT strategy primarily builds a mapping relationship between the source elements and target elements, such as emotion and speaker elements [9].”
Temani, A., & Feng, Y. (2025) · SPE Journal · doi:10.2118/231405-pa · Cited by 1
Enhanced Audio-Visual Speech Synthesis via Multi-Discriminative Learning
Ghosh, S., Zalkow, F., & Jana, N. D. (2025) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2025.3645648 · Cited by 1
“Models such as StarGAN-VC [21] and CycleGAN-VC [22] utilize GANs to learn mappings be-tween different speakers’ voice characteristics without requiring parallel training data.”
Non-Parallel Many-To-Many Voice Conversion Using Local Linguistic Tokens
Wang, C., & Yu, Y. (2021) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp39728.2021.9413540 · Cited by 1
“For GAN based method, CycleGAN-VC[10, 11] incorporates three loss functions: adversarial loss, cycle-consistency loss, and identity mapping loss, to learn the mapping between the source and target spectrum.”
MethodEffects of Sinusoidal Model on Non-Parallel Voice Conversion with Adversarial Learning
Al-Radhi, M. S., Csapó, T. G., & Németh, G. (2021) · Applied Sciences · doi:10.3390/app11167489 · Cited by 1
“One powerful technique that can possibly overcome the weakness of CVAEs is the cycle-consistent generative adversarial network (CycleGAN), which was proposed in [24] using generators, discriminators, and an identity-mapping loss.”
BackgroundTwo-Pathway Style Embedding for Arbitrary Voice Conversion
Xu, X., Shi, L., Chen, J., Chen, X., Lian, J., Lin, P., Zhang, Z., & Hancock, E. R. (2021) · Interspeech (USB) · doi:10.21437/interspeech.2021-506 · Cited by 2
“Recently, generative models have enjoyed great success, such as variational autoencoders (VAEs) [7] and generative adversarial networks (GANs) [8], and these have performed well on non-parallel training corpus [9, 10, 11, 12, 13].”
BackgroundCycleGAN-Based Speech Mode Transformation Model for Robust Multilingual ASR
Tripathi, K., & Rao, K. (2022) · Circuits Systems and Signal Processing · doi:10.1007/s00034-022-02008-4 · Cited by 1
Suda, H., Kotani, G., & SAITO, D. (2022) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2021edp7234 · Cited by 1
Elastic-Wave Reverse Time Migration Random Boundary-Noise Suppression Based on CycleGAN
Xu, G., & He, B. (2022) · Journal of Ocean University of China · doi:10.1007/s11802-022-5171-3 · Cited by 1
“Based on adversarial losses, CycleGAN uses cycle consistency losses (Kaneko and Kameoka, 2018) as a constraint to avoid the conversion of all data in the domain to some data in another domain by the two generators of GAN; as a result, the generated data are consistent with the input itself as much…”
MethodMGVC: A Mask Voice Conversion Using Generating Adversarial Training
Lin, P., Lian, J., & Dai, Y. (2022) · Lecture notes in computer science · doi:10.1007/978-3-031-13829-4_50 · Cited by 1
CycleGAN-based Unpaired Speech Dereverberation
Muckenhirn, H., Safin, A., Erdoğan, H., Quitry, F. D. C., Tagliasacchi, M., Wisdom, S., & Hershey, J. R. (2022) · Interspeech (USB) · doi:10.21437/interspeech.2022-10104 · Cited by 2
Fu, C., Liu, C., Ishi, C. T., & Ishiguro, H. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9979821 · Cited by 1
“Moreover, to better learn the mapping function between non-parallel data distributions, cycle-consistent adversarial network (CycleGAN) [2], [26] and variational autoencoder-generative adversarial network (VAE-GAN) [27] were introduced to the VC task.”
BackgroundMethodLi, W., & Wei, T.-J. (2022) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.23919/apsipaasc55919.2022.9979975 · Cited by 2
“On approach of the VC algorithm is to use generative model, e.g. cycleGAN, starGAN, and Glow , to directly translate voice from one style into another [1, 2, 3, 4, 5, 6].”
MethodZhou, X., Yu, L. H., Niu, F., & Jin, J. (2022) · IEEE Access · doi:10.1109/access.2022.3217466 · Cited by 1
“CycleGAN was introduced to the field of speech conversion by Rafael ferro et al. in [5].”
MethodHimawan, I., Wang, R., Sridharan, S., & Fookes, C. (2022) · IEEE Access · doi:10.1109/access.2022.3226350 · Cited by 1
“Moreover, techniques that do not require transcriptions such as CyleGAN-VC [10], StarGAN-VC [11], and VAW-GAN [12] have incorporated generative adversarial networks that improve voice quality and similarity to the target speaker when a large amount of speech data is employed.”
BackgroundHigh Quality and Similarity One-Shot Voice Conversion Using End-to-End Model
Du, R., & Yao, J. (2022) · doi:10.1145/3577530.3577575 · Cited by 1
“Most VC methods are based on a two-stage processing pipeline, where the acoustic model is responsible for producing the mel-spectrogram, and the vocoder reconstruct the mel-spectrogram into a rawwaveform [4, 5, 14, 15, 19].”
BackgroundMethodAdapting Pretrained Models for Adult to Child Voice Conversion
Sudro, P. N., Ragni, A., & Hain, T. (2023) · European Signal Processing Conference (EUSIPCO) · doi:10.23919/eusipco58844.2023.10289993 · Cited by 1
“However, it is not always possible to collect sufficient quantities of parallel data [3].”
BackgroundDisC-VC: Disentangled and F0-Controllable Neural Voice Conversion
Watanabe, C., & Kameoka, H. (2023) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc58517.2023.10317447 · Cited by 2
“…[10], [11] methods), while others do not (e.g., recognition-synthesis-based [12]–[14], phonetic-posteriorgram-based [15], variational-autoencoder (VAE) [16]- based [17]–[19], generative-adversarial-network-based [20]– [22], flow-based [23], [24], and score-based [25], [26] meth-ods).”
Self-citationFuzzy ASC-GAN: A Fuzzy C-means Audio Similarity CycleGAN on Articulation Disorder Voice Conversion
Huang, Y., Chang, S., Chin, C.-L., Lin, G., Chang, P., Lin, W.-X., Chen, J.-R., Zheng, Z.-H., & Wong, G.-K. (2023) · International conference on Fuzzy Theory and Its Applications · doi:10.1109/ifuzzy60076.2023.10324274 · Cited by 1
MelMAE-VC: Extending Masked Autoencoders to Voice Conversion
Wang, Y., & Gu, Y. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-8145-8_37 · Cited by 1
Improving Voice Style Conversion via Self-attention VAE with Feature Disentanglement
Yuan, H., Li, P., Zhao, G., & Zhang, J. (2024) · Communications in computer and information science · doi:10.1007/978-981-99-9637-7_31 · Cited by 2
A noise-robust voice conversion method with controllable background sounds
Chen, L., Zhang, X., Li, Y., Sun, M., & Chen, W. (2024) · Complex & Intelligent Systems · doi:10.1007/s40747-024-01375-6 · Cited by 1
“The idea of cycle loss was proposed by CycleGAN-VC in [28].”
Xie, A. Z. C. (2024) · Journal of Electrical Systems · doi:10.52783/jes.1841 · Cited by 1
“Non-parallel VC techniques employ GANs [38][39][40] and variational autoencoders (VAEs) [41][42] , with notable success demonstrated by CycleGAN-VC [43] and StarGAN-VC [44][45][46] .”
MethodMaskMel-Prosody-CycleGAN-VC: High-Quality Cross-Lingual Voice Conversion
Yan, S., Chen, S., Xu, Y., & Ke, D. (2024) · Lecture notes in electrical engineering · doi:10.1007/978-981-97-2200-6_2 · Cited by 1
Integrating Noise Classification and Speech Enhancement Model for Hearing Aids
Hung, Y.-H., Chang, Y.-C., Lai, S.-C., Juang, W.-H., Sheu, M., & Lee, J.-D. (2024) · International SoC Design Conference · doi:10.1109/isocc62682.2024.10762526 · Cited by 1
“To compare differences between networks, this paper contrasts the proposed model with the one-to-one voice transformation network CycleGAN [4] .”
Zhou, J., Wang, L., Li, F., Zhang, S., Liu, T., Chen, X., & Niu, H. (2024) · Annual International Conference of the IEEE Engineering in Medicine and Biology Society (EMBC) · doi:10.1109/embc53108.2024.10782916 · Cited by 1
“Utilizing the hierarchical structure of Mandarin prosody, we apply CWT [10] to decompose the EL speech F0 into various frequency scales for feature extraction, which are then fed into CycleGAN [11] for training.”
GPGAN-VC: Enhancing Voice Conversion using Gradient Penalty
Purohit, R. M., Vaghera, D. H., & Patil, H. A. (2024) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc63619.2025.10848670 · Cited by 1
Cycle-GAN Dysarthric-to-Normal Voice Conversion
Chlaikhy, S., Chakhtouna, A., & Adib, A. (2025) · Lecture notes in networks and systems · doi:10.1007/978-3-032-01536-5_63 · Cited by 1
Neural Conversion of Social Attitudes in Speech Signals
Veillon, C. L. M. (2023) · theses.fr (ABES) · doi:10.70675/26188a9az8183z44bez8209z26e91c50ae8f · Cited by 2
Saito, Y., Akuzawa, K., & Tachibana, K. (2020) · IEICE Transactions on Information and Systems · doi:10.1587/transinf.2019edp7297 · Cited by 2
“CycleGAN-VC [26], [27] trains a VC model using non-parallel speech corpora based on the adversarial training considering cyclic-consistency [28].”
MethodFeature Quantization for Many-to-many Voice Conversion
Wang, Y., Xie, Y., Wang, H., & Zhang, Q. (2020) · IEEE International Conference on Information Technology, Big Data and Artificial Intelligence (ICIBA) · doi:10.1109/iciba50161.2020.9276871
Automatic Speech Recognition for the Archive of Ainu Folklores
Matsuura, K., Mimura, M., & Kawahara, T. (2021) · Journal of Natural Language Processing · doi:10.5715/jnlp.28.824
“…CycleGANを利用した教師なし話者適応 3.3節で述べた通り,アイヌ語音声コーパスにはわずか 8名の話者しか含まれておらず,しか も話者ごとのデータ量の分布の偏りが大きい.この結果,4.4節に示した通り,話者オープン条 件における認識性能の大幅な低下が生じた.この問題を「話者スパース問題」と呼ぶことにし, 本節ではこの問題について取り組む. 多くの話者のデータを収集することが困難なのは,低資源言語音声認識において一般に当ては まる.例えば,Grikoコーパス (Boito et al. 2018)は話者 9名,Mboshiコーパス (Rialland et al. 2018)は…”
Speech Generation by Generative Adversarial Network
Chen, Y. (2021) · International Conference on Big Data & Artificial Intelligence & Software Engineering (ICBASE) · doi:10.1109/icbase53849.2021.00086 · Cited by 1
Voice Conversion of Tibetan Amdo Dialect Based on Cycle-Consistent Adversarial Network
Gan, Z., Zhang, M., Wang, R., & Luo, M. (2021) · International Academic Exchange Conference on Science and Technology Innovation (IAECST) · doi:10.1109/iaecst54258.2021.9695907
The UFRJ Entry for the Voice Conversion Challenge 2020
da Costa, V. P., Maia, R., Quintanilha, I. M., Netto, S. L., & Biscainho, L. W. P. (2020) · doi:10.21437/vcc_bc.2020-29
Tagawa, N., Tarashima, S., Yoshida, S., Kawamura, K., Tada, M., & Arai, H. (2020) · The Journal of The Institute of Image Information and Television Engineers · doi:10.3169/itej.74.93
Multiple Speech Mode Transformation using Adversarial Network
Tripathi, K., & Kumar, J. K. (2022) · IEEE Uttar Pradesh Section International Conference on Electrical, Electronics and Computer Engineering (UPCON) · doi:10.1109/upcon56432.2022.9986477
A CONTROLLABLE CROSS-GENDER VOICE CONVERSION FOR SOCIAL ROBOT
Fu, C., Liu, C., Ishi, C. T., & Ishiguro, H. (2022) · International Conference on Affective Computing and Intelligent Interaction Workshops and Demos (ACIIW) · doi:10.1109/aciiw57231.2022.10086038
“In this study, we extracted F0 and spectrogram from the speech as our model’s inputs, inspired by some previous studies [2, 9].”
BackgroundMethodStyle Adaptation Based On Image Processing Methods Using Cyclegan
Popovic, B. Z. (2023) · International Scientific Conference Sinteza · doi:10.15308/sinteza-2023-9-16
“The architecture presented in [16] has been employed, enabling parallel processing of the sequential data.”
MethodSpeech enhancement by CycleGAN using feature map regularization
Popovic, B. Z., & Janev, M. B. (2023) · Journal of Computer and Forensic Sciences · doi:10.5937/jcfs2-45100
Accent-Preserving Voice Conversion between Native-Nonnative Speakers for Second Language Learning
Correa, I., Ueno, S., & Lee, A. (2023) · Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC) · doi:10.1109/apsipaasc58517.2023.10317245
“In [12] a cycle-consistent adversarial network-based non-parallel VC model is proposed for mel-cepstrum conversion.”
BackgroundMethodAziz, S., & Shahnawazuddin, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-48312-7_31
Aziz, S., Pushp, S., & Shahnawazuddin, S. (2023) · Lecture notes in computer science · doi:10.1007/978-3-031-48312-7_32
MISGAN:Mutual Information-Based Unsupervised with Star-GANv2 Voice Conversion
Cui, H., Ping, D., & Wang, K. (2023) · IEEE Smart World Congress (SWC) · doi:10.1109/swc57546.2023.10448606
Zero-Shot Voice Conversion Based on Speaker Embedding Domain Generalization
Lin, Y.-X., Cheng, C.-H., Le, P. T., Huang, B.-J., Chu-Xin, L., Huang, C.-L., & Wang, J. (2023) · RIVF International Conference on Computing and Communication Technologies (RIVF) · doi:10.1109/rivf60135.2023.10471830
Metadata
- Authors
- Kaneko, T., & Kameoka, H.
- Affiliations
- NTT (Japan) (JP)
- Venue
- 2018 26th European Signal Processing Conference (EUSIPCO)
- Publisher
- IEEE
- Published
- 2018-09
- Pages
- 2100-2104
- OpenAlex
- W2902070858
- Cited by
- 358
- References
- 61
- Citations by year
- 2026: 19 · 2025: 36 · 2024: 50 · 2023: 66 · 2022: 58 · 2021: 63 · 2020: 47 · 2019: 18 · 2018: 1
- FWCI
- 13.16
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- European Signal Processing Conference (EUSIPCO) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Health Sciences › Health Professions › Speech and Hearing › Voice and Speech Disorders
- Keywords
- voice conversion, non-parallel voice conversion, CycleGAN, gated convolutional neural network, pseudo data generation, over-smoothing, modulation spectra
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.