Speaker Verification Benchmark

OpenSVBench Leaderboard

Cross-scenario evaluation of speaker verification systems on a fixed public core. The global ranking uses source-balanced scenario-macro EER; lower is better.

25 ranked systems 11 scenarios 26 trial sets 16 source datasets

Submit a system Browse scenarios

Fixed-Core Ranking

Global ranking

Only reviewed, completed submissions with full current-core coverage are included.

Current leader: w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 · JinghanPeng

Rank	System	Submitted by	EER ↓ Scenario macro	minDCF ↓ Scenario macro	Stronger Scenarios	Weaker Scenarios
#1	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2	JinghanPeng	6.0847	0.235010	In-the-wild Distance	Noise/reverb Aging
#2	W2V-BERT 2.0 + Adapter-MFA (cross-device consistency FT + weight soup) ActionPower VoicePrint	ActionPower	7.1907	0.288926	Noise/reverb In-the-wild	Genre shift Accent/dialect
#3	MECT-b2-czur-finetuned MECT-b2-czur-finetuned	czur	7.2276	0.280633	Channel/device Accent/dialect	Aging Noise/reverb
#4	W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2) Wav2Vec2-BERT	pengjy	7.5635	0.368244	Aging Speaking style	Cross-lingual Channel/device
#5	Agora Voiceprint (ResNet293) agora_vp	agora	7.8931	0.297508	Channel/device Distance	Noise/reverb Aging
#6	AntSpeaker/MECT-B2 antspeaker-mect-b2-vb2-full-core	Public checkpoint	8.5599	0.321518	Aging In-the-wild	Genre shift Channel/device
#7	Dual-LoRA W2V-BERT2-Adapter-MFA (TidyVoice) duallora-w2vbert2-tidyvoice-full-core	sgqt	9.0606	0.342596	Cross-lingual Speaking style	Genre shift Accent/dialect
#8	palabraai/redimnet2-b6-vb2+vox2 redimnet2-b6-vb2vox2-lm-full-core	Public checkpoint	9.1442	0.324160	Noise/reverb Aging	Genre shift Accent/dialect
#9	wespeaker/w2vbert2 vb2-w2vbert	Public checkpoint	9.9604	0.360767	Noise/reverb Speaking style	Genre shift Channel/device
#10	palabraai/redimnet2-b6 redimnet2	Public checkpoint	10.5277	0.419432	Noise/reverb Aging	Genre shift In-the-wild
#11	wespeaker/samresnet100 vb2-sam100	Public checkpoint	10.5780	0.390239	Aging Noise/reverb	Accent/dialect Channel/device
#12	wespeaker/res152-voxceleb wespeaker_resnet152	Public checkpoint	11.2275	0.442748	Speaking style Aging	Genre shift Channel/device
#13	wespeaker/res293-voxceleb wespeaker_resnet293	Public checkpoint	11.3130	0.439607	Aging Cross-lingual	Genre shift In-the-wild
#14	iic/eres2netv2-zh eresv2_zh	Public checkpoint	11.5449	0.484961	Genre shift Channel/device	Noise/reverb Overlap
#15	wespeaker/res34-voxceleb wespeaker_resnet34_voxceleb	Public checkpoint	12.0516	0.477054	Aging Speaking style	Channel/device Short-duration
#16	iic/eres2net-en eres_en	Public checkpoint	12.0955	0.472802	Noise/reverb Overlap	Genre shift Short-duration
#17	wespeaker/campplus-voxceleb campplus_wespeaker	Public checkpoint	13.4165	0.520675	Aging Overlap	Genre shift Channel/device
#18	speechbrain/ecapa ecapa	Public checkpoint	13.5461	0.543660	Noise/reverb Accent/dialect	Cross-lingual Short-duration
#19	wespeaker/ecapa1024-voxceleb wespeaker_ecapa1024	Public checkpoint	14.0835	0.534111	Aging Speaking style	Genre shift Short-duration
#20	iic/eres2net-large-3dspeaker eres2net_large_3dspeaker	Public checkpoint	14.1682	0.593516	Channel/device Distance	Aging In-the-wild
#21	wespeaker/ecapa512-voxceleb wespeaker_ecapa512	Public checkpoint	14.3120	0.541141	Aging Cross-lingual	Genre shift Channel/device
#22	wespeaker/res34-cnceleb wespeaker_r34	Public checkpoint	14.4910	0.607964	Genre shift Short-duration	Channel/device Distance
#23	iic/campplus campplus	Public checkpoint	15.1677	0.590178	Genre shift Channel/device	Noise/reverb Accent/dialect
#24	microsoft/wavlm-base-plus-sv wavlm_base	Public checkpoint	25.1577	0.875090	Aging In-the-wild	Genre shift Distance
#25	microsoft/unispeech-sat-base-plus-sv unispeech_sat_base_plus	Public checkpoint	25.6711	0.890736	Noise/reverb Short-duration	Accent/dialect Channel/device

Diagnostic Views

Scenario overview

Use scenario rankings to inspect deployment-specific strengths and failure modes hidden by the aggregate score.

Open full scenario index

Scenario	Evaluation Focus	Trial Sets	Current Leader
Accent / Dialect Robustness Accent/dialect	Accent and dialect variation	2	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
Aging Robustness Aging	Speaker aging and time gaps	2	W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2) pengjy
Channel / Device Robustness Channel/device	Device and channel variation	2	MECT-b2-czur-finetuned czur
Cross-Lingual Robustness Cross-lingual	Language mismatch	1	Dual-LoRA W2V-BERT2-Adapter-MFA (TidyVoice) sgqt
Distance Robustness Distance	Distance mismatch	4	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
Genre-Shift Robustness Genre shift	Source-genre variation	1	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
In-The-Wild Robustness In-the-wild	Open-domain media speech	4	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
Noise / Reverb Robustness Noise/reverb	Noise and reverberation	1	palabraai/redimnet2-b6-vb2+vox2 Public checkpoint
Overlap Robustness Overlap	Overlapping speakers	2	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
Short-Duration Robustness Short-duration	Short-duration speech	4	w2v-bert2.0 + cc + ecapa-tdnn + astp + sf2 JinghanPeng
Speaking-Style Robustness Speaking style	Speaking style shift	3	W2V-BERT 2.0 + MHFA (VoxCeleb2 + VoxBlink2) pengjy

Global Reach

Visitor map

Country-level activity, updated hourly.

Country-level world map for OpenSVBench page visits.

Total visits: 15,468
Countries / regions: 96
Last 30 days: 15,468