VoxCeleb: A Large-Scale Speaker Identification Dataset
Open accessThe DOI matches its Crossref record; 10 of the 10 references it cites were found in a registry or an index; none of the 16 sources its claims cite is checked for support yet.
Abstract
Most existing datasets for speaker identification contain samples obtained under quite constrained conditions, and are usually hand-annotated, hence limited in size. The goal of this paper is to generate a large scale text-independent speaker identification dataset collected 'in the wild'. We make two contributions. First, we propose a fully automated pipeline based on computer vision techniques to create the dataset from open-source media. Our pipeline involves obtaining videos from YouTube; performing active speaker verification using a two-stream synchronization Convolutional Neural Network (CNN), and confirming the identity of the speaker using CNN based facial recognition. We use this pipeline to curate VoxCeleb which contains hundreds of thousands of 'real world' utterances for over 1,000 celebrities. Our second contribution is to apply and compare various state of the art speaker identification techniques on our dataset to establish baseline performance. We show that a CNN based architecture obtains the best performance for both identification and verification.
Overview
CoChat Summary
TL;DR
The paper introduces VoxCeleb, a fully automated, large-scale speaker identification dataset built from YouTube videos “in the wild,” with 1,251 celebrities and over 100,000 utterances. It also benchmarks speaker recognition methods and reports that a CNN performs best on both identification and verification.
Key findings
- VoxCeleb contains over 100,000 utterances for 1,251 celebrities; the dataset is gender balanced, with 55% male speakers.
- The automated collection pipeline uses active speaker verification and face verification and was tuned to be highly precise: precision = 1.000 for both stages, with recall = 0.613 for active speaker verification and 0.726 for face verification.
- For speaker identification, the CNN achieves 80.5% top-1 and 92.1% top-5 accuracy, compared with 49.0% / 56.6% for I-vectors + SVM and 60.8% / 75.6% for I-vectors + PLDA + SVM.
- For speaker verification, the best reported result is CNN + Embedding with C min det = 0.71 and EER = 7.8%, better than I-vectors + PLDA (0.73, 8.8%) and GMM-UBM (0.80, 15.0%).
Methods
- Built VoxCeleb from YouTube videos using a fully automated multi-stage pipeline: candidate POI list, video download, face tracking, active speaker verification via SyncNet, and face verification via VGG Face CNN.
- Dataset scale/statistics reported: 1,251 POIs, 690 male POIs, 36 / 18 / 8 videos per POI (max/avg/min), 250 / 123 / 45 utterances per POI (max/avg/min), and 145.0 / 8.2 / 4.0 s utterance length (max/avg/min).
- Trained and evaluated a CNN on spectrograms with minimal preprocessing; identification used a 1,251-way classification setup, while verification compared I-vectors/PLDA, GMM-UBM, and CNN embedding approaches.
Limitations
- The paper states that the pipeline is designed to be conservative, and that this means “a lot of the downloaded videos” are discarded to reduce label errors.
- The paper notes that the collection starts from a list of 2,622 identities from VGG Face and uses YouTube search with “interview” appended, which is an evident source of selection bias toward publicly available, interview-like media.
- The dataset is described as coming from open-source media / YouTube, so it is limited to speakers and contexts available there; this is evident from the text, though not stated as a formal limitation.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
0 Valid · 1 Not valid · 12 Not checked · 37 Own results and interpretations
In the paper
“| Face verification | [30] | 1.000 | 0.726 |”
What the sources say
- Not checked yetcites [30] [30] O. M. Parkhi, A. Vedaldi, and A. Zisserman, 'Deep face recognition,' in Proceedings of the British Machine Vision Conference , 2015. -
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“Unlike the SITW dataset, both audio and video for each speaker is released.”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
In the paper
“| Active speaker verification | [31] | 1.000 | 0.613 |”
What the sources say
- Not checked yetcites [31] [31] P. Chakravarty and T. Tuytelaars, 'Cross-modal supervision for learning active speaker detection in video,' arXiv preprint arXiv:1603.08907 , 2016. -
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper states — it cites no source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“reduces the number of parameters from 319M in VGG-M to 67M in our network”
plainstates a cause- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
A result the paper reports from its own data — it cites no other source for it, so there is none to check. The quote shows it is in the paper.
In the paper
“| # of videos per POI | 36 / 18 / 8 |”
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: unclear — It is unclear whether the paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
15 more claims
Create a free account to see all 50 claims and to verify the 16 references still open.
- Probably not a substantial claim. 9
- It may state two separate claims as one. 23
- The quote only partly states this claim. 5
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?
The cited source says what the claim attributes to it.
In the citing paper
“Meanwhile, the emergence of large-scale VSA datasets [28, 29, 30, 31, 32] promoted the further development of deep learning based VSA research.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [31]
In the citing paper
“Although it is mainly built for speaker identification, it also can be used for VSG.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [31]
In the citing paper
“It contains over 100k utterances from 1,251 celebrities.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [31]
In the citing paper
“VoxCeleb1 [31] is a large-scale text-independent audio-visual dataset collected from open-source YouTube media.”
From Deep learning for visual speech analysis: A survey (2024) · cites this paper as [31]
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
Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S., & Szegedy, C. (2015) · arXiv · doi:10.48550/arxiv.1502.03167 · Cited by 27,818 · cited 1×
“The network is trained using batch normalisation [40] and all hyper-parameters (e.”
MethodBatch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
Ioffe, S. (2024) · arXiv (Cornell University) · doi:10.57702/o9raffed · Cited by 14,214
Deep Residual Learning for Image Recognition
He, K., Zhang, X., Ren, S., & Sun, J. (2015) · arXiv · doi:10.48550/arxiv.1512.03385 · Cited by 4,196
One millisecond face alignment with an ensemble of regression trees
Kazemi, V., & Sullivan, J. (2014) · 2014 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2014.241 · Cited by 2,831 · cited 1×
“Facial landmark positions are detected for each face detection using the regression tree based method of [33].”
MethodCNN architectures for large-scale audio classification
Hershey, S., Chaudhuri, S., Ellis, D. P. W., Gemmeke, J. F., Jansen, A., Moore, R. C., Plakal, M., Platt, D., Saurous, R. A., Seybold, B., et al. (2017) · 2017 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2017.7952132 · Cited by 2,023
MatConvNet - Convolutional Neural Networks for MATLAB
Vedaldi, A., & Lenc, K. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1412.4564 · Cited by 578 · cited 1×
“Our implementation is based on the deep learning toolbox MatConvNet [39] and trained on a NVIDIA TITAN X GPU.”
MethodCNN Architectures for Large-Scale Audio ClassificationPreprint
Hershey, S., Chaudhuri, S., Ellis, D. P. W., Gemmeke, J. F., Jansen, A., Moore, R. C., Plakal, M., Platt, D., Saurous, R. A., Seybold, B., et al. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1609.09430 · Cited by 217 · cited 1×
“This has led to a very recent shift from hand-crafted features to the domain of deep CNNs which can be applied to higher dimensional inputs [13, 14] and for speaker iden-tification [15].”
BackgroundCross-Modal Supervision for Learning Active Speaker Detection in Video
Chakravarty, P., & Tuytelaars, T. (2016) · Lecture notes in computer science · doi:10.1007/978-3-319-46454-1_18 · Cited by 70 · cited 2×
“Precision-recall curves for both tasks on their respective benchmark datasets [30, 31] are shown in Figure 1 right, and the values at the operating point are given in Table 3.”
BackgroundCross-modal Supervision for Learning Active Speaker Detection in VideoPreprint
Chakravarty, P., & Tuytelaars, T. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1603.08907 · Cited by 1 · cited 2×
“Precision-recall curves for both tasks on their respective benchmark datasets [30, 31] are shown in Figure 1 right, and the values at the operating point are given in Table 3.”
Background
Research works printed but not linked: 28
- [2][2] A. Krizhevsky, I. Sutskever, and G. E. Hinton, 'ImageNet classification with deep convolutional neural networks,' in Advances in Neural Information Processing Systems , pp. 1106-1114, 2012. -
- [5][5] M. McLaren, L. Ferrer, D. Castan, and A. Lawson, 'The speakers in the wild (SITW) speaker recognition database,' INTERSPEECH , vol. 2016, 2016. -
- [6][6] O. Russakovsky, J. Deng, H. Su, J. Krause, S. Satheesh, S. Ma, S. Huang, A. Karpathy, A. Khosla, M. Bernstein, A. Berg, and F. Li, 'Imagenet large scale visual recognition challenge,' International Journal of Computer Vision , 2015. -
- [7][7] D. A. Reynolds, T. F. Quatieri, and R. B. Dunn, 'Speaker verification using adapted gaussian mixture models,' Digital signal processing , vol. 10, no. 1-3, pp. 19-41, 2000. -
- [8][8] D. A. Reynolds and R. C. Rose, 'Robust text-independent speaker identification using gaussian mixture speaker models,' IEEE transactions on speech and audio processing , vol. 3, no. 1, pp. 7283, 1995. -
- [10][10] N. Dehak, P. J. Kenny, R. Dehak, P. Dumouchel, and P. Ouellet, 'Front-end factor analysis for speaker verification,' IEEE Transactions on Audio, Speech, and Language Processing , vol. 19, no. 4, pp. 788-798, 2011. -
- [11][11] U. H. Yapanel, X. Zhang, and J. H. Hansen, 'High performance digit recognition in real car environments.,' in INTERSPEECH , 2002. -
- [12][12] J. H. Hansen, R. Sarikaya, U. H. Yapanel, and B. L. Pellom, 'Robust speech recognition in noise: an evaluation using the spine corpus.,' in INTERSPEECH , pp. 905-908, 2001. -
- [13][13] T. N. Sainath, R. J. Weiss, A. W. Senior, K. W. Wilson, and O. Vinyals, 'Learning the speech front-end with raw waveform CLDNNs,' in INTERSPEECH , pp. 1-5, 2015. -
- [15][15] Y. Lukic, C. Vogt, O. D¨ urr, and T. Stadelmann, 'Speaker identification and clustering using convolutional neural networks,' in IEEE 26th International Workshop on Machine Learning for Signal Processing (MLSP) , pp. 1-6, IEEE, 2016. -
- [16][16] D. van der Vloed, J. Bouten, and D. A. van Leeuwen, 'NFIFRITS: a forensic speaker recognition database and some first experiments,' in The Speaker and Language Recognition Workshop , 2014. -
- [17][17] J. Hennebert, H. Melin, D. Petrovska, and D. Genoud, 'POLYCOST: a telephone-speech database for speaker recognition,' Speech communication , vol. 31, no. 2, pp. 265-270, 2000. -
- [18][18] J. B. Millar, J. P. Vonwiller, J. M. Harrington, and P. J. Dermody, 'The Australian national database of spoken language,' in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing , vol. 1, pp. I-97, IEEE, 1994. -
- [19][19] J. S. Garofolo, L. F. Lamel, W. M. Fisher, J. G. Fiscus, and D. S. Pallett, 'DARPA TIMIT acoustic-phonetic continous speech corpus CD-ROM. NIST speech disc 1-1.1,' NASA STI/Recon technical report , vol. 93, 1993. -
- [21][21] R. Woo, A. Park, and T. J. Hazen, 'The MIT Mobile Device Speaker Verification Corpus: Data collection and preliminary experiments,' The Speaker and Language Recognition Workshop , 2006. -
- [23][23] A. Janin, D. Baron, J. Edwards, D. Ellis, D. Gelbart, N. Morgan, B. Peskin, T. Pfau, E. Shriberg, A. Stolcke, et al. , 'The ICSI meeting corpus,' in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing , vol. 1, IEEE, 2003. -
- [24][24] I. McCowan, J. Carletta, W. Kraaij, S. Ashby, S. Bourban, M. Flynn, M. Guillemot, T. Hain, J. Kadlec, V. Karaiskos, et al. , 'The AMI meeting corpus,' in International Conference on Methods and Techniques in Behavioral Research , vol. 88, 2005. -
- [25][25] P. Bell, M. J. Gales, T. Hain, J. Kilgour, P. Lanchantin, X. Liu, A. McParland, S. Renals, O. Saz, M. Wester, et al. , 'The MGB challenge: Evaluating multi-genre broadcast media recognition,' in IEEE Workshop on Automatic Speech Recognition and Understanding , pp. 687-693, IEEE, 2015. -
- [27][27] J. J. Godfrey, E. C. Holliman, and J. McDaniel, 'Switchboard: Telephone speech corpus for research and development,' in Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing , vol. 1, pp. 517-520, IEEE, 1992. -
- [28][28] W. M. Fisher, G. R. Doddington, and K. M. Goudie-Marshall, 'The DARPA speech recognition research database: specifications and status,' in Proc. DARPA Workshop on speech recognition , pp. 93-99, 1986. -
- [30][30] O. M. Parkhi, A. Vedaldi, and A. Zisserman, 'Deep face recognition,' in Proceedings of the British Machine Vision Conference , 2015. -
- [32][32] D. E. King, 'Dlib-ml: A machine learning toolkit,' The Journal of Machine Learning Research , vol. 10, pp. 1755-1758, 2009. -
- [34][34] J. S. Chung and A. Zisserman, 'Lip reading in the wild,' in Proceedings of the Asian Conference on Computer Vision , 2016. -
- [35][35] M. Everingham, J. Sivic, and A. Zisserman, 'Taking the bite out of automatic naming of characters in TV video,' Image and Vision Computing , vol. 27, no. 5, 2009. -
- [36][36] J. S. Chung and A. Zisserman, 'Out of time: automated lip sync in the wild,' in Workshop on Multi-view Lip-reading, ACCV , 2016. -
- [37][37] K. Chatfield, K. Simonyan, A. Vedaldi, and A. Zisserman, 'Return of the devil in the details: Delving deep into convolutional nets,' in Proceedings of the British Machine Vision Conference , 2014. -
- [38][38] S. Chopra, R. Hadsell, and Y. LeCun, 'Learning a similarity metric discriminatively, with application to face verification,' in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , vol. 1, pp. 539-546, IEEE, 2005. -
- [41][41] S. Ioffe, 'Probabilistic linear discriminant analysis,' in Proceedings of the European Conference on Computer Vision , pp. 531542, Springer, 2006.
Cited by
X-Vectors: Robust DNN Embeddings for Speaker Recognition
Snyder, D., Garcia-Romero, D., Sell, G., Povey, D., & Khudanpur, S. (2018) · 2018 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) · doi:10.1109/icassp.2018.8461375 · Cited by 2,620
“5 we incorporate audio from the new VoxCeleb dataset [19] into both extractor and PLDA training lists.”
BackgroundMethodVoxCeleb2: Deep Speaker Recognition
Chung, J. S., Nagrani, A., & Zisserman, A. (2018) · Interspeech 2018 · doi:10.21437/interspeech.2018-1929 · Cited by 2,256
Self-citationWavLM: Large-Scale Self-Supervised Pre-Training for Full Stack Speech Processing
Chen, S., Wang, C., Chen, Z., Wu, Y., Liu, S., Chen, Z., Li, J., Kanda, N., Yoshioka, T., Xiao, X., et al. (2022) · IEEE Journal of Selected Topics in Signal Processing · doi:10.1109/jstsp.2022.3188113 · Cited by 1,926
“Compared with the Fbank feature, the representations from WavLM Large achieve over 35% relative EER improvement on all three trials for the VoxCeleb1 evaluation set.”
BackgroundInfluentialDesplanques, B., Thienpondt, J., & Demuynck, K. (2020) · Interspeech 2020 · doi:10.21437/interspeech.2020-2650 · Cited by 1,568
Speaker Recognition from Raw Waveform with SincNet
Ravanelli, M., & Bengio, Y. (2018) · 2018 IEEE Spoken Language Technology Workshop (SLT) · doi:10.1109/slt.2018.8639585 · Cited by 896
Metadata
- Authors
- Nagrani, A., Chung, J. S., & Zisserman, A.
- Venue
- Interspeech 2017
- Publisher
- ISCA
- Published
- 2017-08-20
- Pages
- 2616-2620
- ISSN
- 2308-457X (ISSN-L)
- OpenAlex
- W2726515241
- Cited by
- 2,210
- References
- 4
- Citations by year
- 2026: 144 · 2025: 280 · 2024: 313 · 2023: 361 · 2022: 330 · 2021: 319 · 2020: 265 · 2019: 151 · 2018: 47
- FWCI
- 56.51
- Type
- conference paper
- Language
- en
- Open access
- green
- arxiv.org
- Locations
- Interspeech (USB) (publishedVersion)
- arXiv (Cornell University) (submittedVersion)
- arXiv (Cornell University)
- Funders
- Engineering and Physical Sciences Research Council (EP/M013774/1)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Face recognition and analysis
- Keywords
- speaker identification, speaker verification, VoxCeleb, text-independent speaker recognition, automated dataset generation, two-stream CNN, face recognition, YouTube videos
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.
A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
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.