README.md

---
model-index:
- name: BrightoSV-SpeakerVerification-V1.2
  results:
  - task:
      type: speaker-verification
      name: Speaker Verification
    metrics:
    - type: EER
      value: 1.184
      name: Equal Error Rate (%) - Bank-Grade 4s / 5-Enroll
    - type: EER
      value: 1.751
      name: Equal Error Rate (%) - Consumer 2s / 5-Enroll
    - type: other
      value: 3.18
      name: FRR @ FAR=0.1% (%) - Bank-Grade 4s / 5-Enroll
    - type: other
      value: 6.65
      name: FRR @ FAR=0.01% (%) - Bank-Grade 4s / 5-Enroll
    - type: other
      value: 5.3463
      name: Tail Gap 5% - Bank-Grade 4s / 5-Enroll
tags:
- audio
- speaker-verification
- voice-biometrics
- bank-grade
- biometric-security
- pytorch
- multilingual
- language-agnostic
- commercial-sota
- as-norm
language:
- vi
- en
- zh
- ko
- de
- fr
- ja
- ar
- nl
- id
license: other
---

# 🏆 BrightoSV Speaker Verification V1.2

<div align="center">

### **COMMERCIAL SOTA • GLOBAL RELEASE**

**Bank-Grade Voice Identity Verification**
*(Xác thực Định danh Giọng nói cấp Ngân hàng)*

<div align="center">
<p>
<img src="https://huggingface.co/datasets/thusinh1969/musan/resolve/main/BrighTO-SV-V1.2-E7-VN.jpg?download=true" width="100%" alt="BrightoSV V1.5 SOTA Performance"/>
</p>
</div>

</div>

---

**BrightoSV Speaker Verification V1.2** is a **Commercial SOTA** voiceprint verification system, engineered for **offline, on-premise deployment** in banking, government, healthcare, and enterprise sectors. It answers the fundamental question: **"Is this person who they claim to be?"**

*(BrightoSV Speaker Verification V1.2 là hệ thống xác thực giọng nói SOTA thương mại, được thiết kế cho **triển khai nội bộ, hoàn toàn offline** trong ngân hàng, chính phủ, y tế và doanh nghiệp.)*

Trained on a **massive multilingual corpus with 22,000++ speakers** spanning **9+ languages** with **600,000+ speaker cohort** for score normalization, evaluated on **20 million scored pairs** — delivering statistically ironclad results at extreme operating points.

---

## 🏆 Key Performance Indicators

### 🏦 Bank-Grade Benchmarks (4-Second QA Gate, 4s/2s Windows)

Evaluated on **10,000,000 positive + 10,000,000 negative pairs** with strict bank-grade QA (Audio ≥ 4s, SNR ≥ 10dB, Speech Ratio ≥ 15%):

| Metric *(Chỉ số)* | 3-Enrollment | 5-Enrollment | Significance *(Ý nghĩa)* |
| --- | --- | --- | --- |
| **EER** | **1.478%** | **1.184%** 👑 | 🏆 **Commercial SOTA.** Sub-1.2% on diverse multilingual test set. *(Tỷ lệ lỗi cân bằng dưới 1.2%).* |
| **FRR @ FAR=0.1%** | **4.31%** | **3.18%** 👑 | ✅ **Bank-Grade Achieved.** Under 5% at 1-in-1,000 impostor protection. *(Đạt chuẩn ngân hàng tại FAR 0.1%).* |
| **FRR @ FAR=0.01%** | **8.60%** | **6.65%** | 🔒 **Maximum Security.** 1-in-10,000 impostor protection for high-value transactions. *(Bảo vệ 1/10,000 cho giao dịch giá trị cao).* |
| **Tail Gap 5%** | **+4.4628** | **+5.3463** 👑 | 🛡️ **Full Separation.** 95th percentile genuine scores fully above worst impostor region. *(Phân tách hoàn toàn vùng đuôi).* |
| **Latency** ⚡ | — | **< 60ms** | Real-time processing on consumer GPU. *(Xử lý thời gian thực).* |

### 📱 Extended Coverage Benchmarks (2-Second QA Gate, 2s/1s Windows)

For scenarios requiring shorter audio — mobile apps, call centers, IoT:

| Metric *(Chỉ số)* | 3-Enrollment | 5-Enrollment | Significance *(Ý nghĩa)* |
| --- | --- | --- | --- |
| **EER** | **2.362%** | **1.751%** 👑 | 🏆 **Best-in-Class.** Strong accuracy even with 2-second audio. *(Độ chính xác cao ngay cả với audio 2 giây).* |
| **FRR @ FAR=0.1%** | **8.02%** | **6.16%** | ✅ **Practical Consumer UX.** Manageable retry rate for mobile. *(Tỷ lệ xác thực lại phù hợp ứng dụng di động).* |
| **FRR @ FAR=0.01%** | **15.91%** | **12.58%** | 🔒 **High Security on Short Audio.** Viable for multi-factor deployments. *(Bảo mật cao trên audio ngắn).* |
| **Tail Gap 5%** | **+2.3151** | **+3.1082** | 🛡️ **Positive Separation.** Genuine and impostor tails clearly separated. *(Vùng đuôi tách biệt rõ ràng).* |

---

## 📊 Evaluation Methodology — Statistical Rigor

Unlike many speaker verification benchmarks that report on small test sets, BrightoSV V1.2 is evaluated at **industrial scale**:

| Aspect | Detail |
| --- | --- |
| **Positive pairs** | **10,000,000** (same-speaker, cross-utterance) |
| **Negative pairs** | **10,000,000** (different-speaker, balanced 1:1) |
| **Total scored pairs** | **20,000,000** |
| **Unique speakers** | **3,900+** in evaluation set |
| **Multi-enrollment** | 3 and 5 enrollment utterances, mean-aggregated |
| **Score normalization** | **AS-NORM** with **600,000+ speaker cohort** |
| **Quality fusion** | **QMF** (Quality Metric Fusion) — compensates for speaker-specific and duration offsets |
| **QA gates** | 5-gate bank-grade: Duration, Clipping, RMS Energy, Speech Ratio, SNR |

This scale ensures that operating points at **FAR=0.01%** (1 in 10,000) are backed by actual counts of **1,000 impostor threshold crossings**, not statistical extrapolation.

*(Quy mô này đảm bảo các chỉ số tại FAR=0.01% được xác thực bởi 1,000 mẫu vượt ngưỡng thực tế, không phải ngoại suy thống kê.)*

---

## 🌍 Multilingual Training — Global Voice Coverage

BrightoSV V1.2 is trained on a **large-scale multilingual corpus** ensuring language-agnostic voiceprint extraction:

| Language | Coverage | Notes |
| --- | --- | --- |
| 🇻🇳 **Vietnamese** | ★★★★★ | **Primary language.** Extensive dialect coverage (Northern, Central, Southern). *(Ngôn ngữ chính, bao phủ đầy đủ phương ngữ Bắc, Trung, Nam).* |
| 🇬🇧 **English** | ★★★★★ | **Primary languages** Multiple accents (US, UK, AU, Indian, Singapore) |
| 🇨🇳 **Chinese** | ★★★★★ | **Primary languages** Mandarin and regional variants |
| 🇰🇷 **Spanish** | ★★★★☆ | Native speaker corpus |
| 🇩🇪 **German** | ★★★★☆ | European language coverage |
| 🇫🇷 **French** | ★★★★☆ | Including African French variants |
| 🇳🇱 **Dutch** | ★★★★☆ | European language coverage |
| 🇯🇵 **Japanese** | ★★★★☆ | Native speaker corpus |
| 🇰🇷 **Korean** | ★★★★☆ | Native speaker corpus |
| 🇸🇦 **Arabic** | ★★★★☆ | Multiple dialect coverage |

**Key principle:** Speaker identity is carried by vocal tract shape, pitch dynamics, and articulatory patterns — these are **language-independent**. A speaker can enroll in Vietnamese and verify in English. The model extracts *who is speaking*, not *what is being said*.

*(Nguyên tắc then chốt: Định danh người nói mang tính phổ quát, không phụ thuộc ngôn ngữ. Người dùng có thể đăng ký bằng tiếng Việt và xác thực bằng tiếng Anh.)*

---

## 🛡️ Robustness — Augmentation & Real-World Resilience

The model is battle-tested against real-world audio degradation through **comprehensive augmentation** during training:

### 🔊 Noise Resilience

| Category | Examples | Goal |
| --- | --- | --- |
| 🏙️ **Urban** | Street noise, sirens, traffic, construction | On-the-go verification |
| 🏠 **Domestic** | TV/radio background, appliances, children | Work-from-home reliability |
| 🗣️ **Babble** | Crowd noise, overlapping speakers, cafeteria | The hardest scenario — solved |
| ⛈️ **Natural** | Wind, rain, thunder | Outdoor stability |
| 🐾 **Biological** | Coughing, sneezing, baby crying | Disentangle speaker from artifacts |

### 📡 Channel & Codec Resilience

| Codec / Channel | Simulation |
| --- | --- |
| **GSM / AMR** | Mobile telephony compression |
| **VoIP (Zalo, WhatsApp)** | Internet calling artifacts |
| **MP3 / AAC / OGG** | Lossy compression at various bitrates |
| **Microphone variance** | Laptop, phone, headset, far-field |
| **Room acoustics** | Reverb, echo, room impulse response |

### 🎛️ SpecAugment

Time and frequency masking applied during training to prevent overfitting to specific spectral patterns, forcing the model to learn robust speaker representations from partial information.

**Result:** The model maintains accuracy down to **10 dB SNR** — equivalent to speaking in a moderately noisy café. Below 10 dB, the QA gate rejects the audio before inference, protecting against unreliable decisions.

---

## 🎯 Production Deployment

### Three Security Levels

| Level | Min Audio | Enrollment | FAR Options | Use Case |
| --- | --- | --- | --- | --- |
| 🏦 **`bank_strict`** | ≥ 4.0s | **5 samples** | 0.1% / 0.01% | High-value banking, government *(Ngân hàng giá trị cao)* |
| 🏛️ **`bank_flex`** | ≥ 4.0s | **3 samples** | 0.1% / 0.01% | Standard banking, telecom *(Ngân hàng tiêu chuẩn)* |
| 📱 **`consumer`** | ≥ 2.0s | **3 samples** | 0.1% / 0.01% | Mobile apps, call centers, IoT *(Ứng dụng di động)* |

### Scoring Pipeline

```
Audio → QA Gate (5 checks) → Windowed Embedding Extraction → AS-NORM (600K cohort) → QMF → Decision
```

| Component | Detail |
| --- | --- |
| **Embedding** | 512-dimensional voiceprint |
| **Score normalization** | AS-NORM with top-300 cohort matching |
| **Quality fusion** | Cohort Mean Fusion + Duration compensation |
| **Multi-enrollment** | Mean-aggregated across sessions |
| **Storage** | ~2 KB per enrolled speaker |

### QA Gate — Mandatory Pre-Check

All audio **MUST** pass bank-grade QA before inference:

| Check | Threshold | Purpose |
| --- | --- | --- |
| Duration | ≥ 2s or ≥ 4s | Sufficient speech content |
| Clipping | < 0.1% | No distorted audio |
| RMS Energy | −45 to −5 dBFS | Proper recording level |
| Speech Ratio | ≥ 15% | Actual speech, not silence |
| SNR | ≥ 10 dB | Acceptable noise level |

*Without QA gate, tail performance degrades significantly. QA is mandatory for production.*
*(Không có QA gate, hiệu suất vùng đuôi giảm đáng kể. QA là bắt buộc.)*

---

## 🆚 Performance Context

### Why These Numbers Matter

| Metric | BrightoSV V1.2 (4s/5e) | What It Means |
| --- | --- | --- |
| **EER 1.184%** | For every 1,000 verification attempts, ~12 are errors (combined false accept + false reject) |
| **FRR 3.18% @ FAR 0.1%** | At 1-in-1,000 impostor protection: only 3.2% genuine users need to retry |
| **FRR 6.65% @ FAR 0.01%** | At 1-in-10,000 impostor protection: only 6.7% genuine users need to retry |
| **Tail Gap +5.35** | The hardest 5% of genuine speakers are still **5.35 score units above** the strongest impostor region |

### Scale Comparison

| Aspect | Typical Academic Eval | BrightoSV V1.2 |
| --- | --- | --- |
| Test pairs | ~100K–500K | **20,000,000** |
| Speakers | ~100–500 | **3,900+** |
| Score normalization | Often none | **AS-NORM (600K cohort)** |
| Multi-enrollment | Rarely tested | **3-enroll and 5-enroll** |
| QA gates | Rarely applied | **5-gate bank-grade** |
| Languages | Usually 1–2 | **9+ languages** |

---

## ⚙️ Technical Specifications

| Specification | Value |
| --- | --- |
| **Model Version** | V1.2 (Commercial SOTA) |
| **Parameters** | **316M** (High-Capacity Self-Supervised Backbone) |
| **Embedding Dimension** | **512** |
| **Input Sample Rate** | 16kHz (Auto-resampling supported) |
| **Input Formats** | WAV, FLAC, MP3, OGG, M4A |
| **Output** | 512D L2-normalized embedding |
| **Backends** | PyTorch, ONNX, HuggingFace |

### 🚀 Hardware & Performance

| Specification | Value |
| --- | --- |
| **GPU Support** | NVIDIA T4, A10, A100, H100, L4 |
| **CPU Support** | Intel Xeon, AMD EPYC (via ONNX) |
| **Inference Latency** | **< 60ms** (GPU) / **< 400ms** (CPU ONNX) |
| **Model Size** | ~1.2 GB |
| **Batch Processing** | Supported |
| **Deployment** | Fully offline after initial download |

---

## 🌍 Application Scenarios

| Sector | Use Case | Recommended Level |
| --- | --- | --- |
| 🏦 **Banking & Finance** | Wire transfers, Voice Banking, Phone Banking | `bank_strict` |
| 🆔 **eKYC** | Customer onboarding, Remote identity verification | `bank_flex` |
| 🪙 **Crypto & FinTech** | Wallet protection, Transaction authorization | `bank_strict` |
| ✈️ **National Security** | Border control, Immigration screening | `bank_strict` |
| 🎧 **Call Centers** | Caller identity verification, Fraud prevention | `bank_flex` |
| 🏥 **Healthcare** | Patient identity, Telemedicine authentication | `bank_flex` |
| 📱 **Consumer Apps** | Voice login, Smart home, Voice assistants | `consumer` |

---

## 🔒 Privacy & Security

| Aspect | Implementation |
| --- | --- |
| **Audio Retention** | **Zero.** Audio processed in RAM, immediately discarded. *(Không lưu audio).* |
| **Voiceprint** | 512 numbers. **Non-reversible** — cannot reconstruct voice. *(Không thể tái tạo giọng nói).* |
| **Deployment** | On-premise or private cloud. No external calls. *(Triển khai nội bộ, không gọi ra ngoài).* |
| **Compliance** | **GDPR**, **PDPA**, **PCI-DSS** ready |
| **Data Sovereignty** | 100% local processing. Your data never leaves your infrastructure. |

---

## 🤝 Combined with Anti-Spoofing

For maximum security, deploy BrightoSV Speaker Verification **alongside** [BrightoSV Anti-Spoofing V1.5](https://huggingface.co/hatto-ai/BrightoSV-AntiSpoof-V1.5-Special-Edition):

```
Audio → Anti-Spoof Check (Is this a real voice?) → Speaker Verify (Is this the right person?) → Decision
```

| Layer | Model | Purpose |
| --- | --- | --- |
| **Layer 1** | Anti-Spoof V1.5 | Reject deepfakes, replay attacks, TTS | 
| **Layer 2** | Speaker Verify V1.2 | Confirm speaker identity |

This dual-layer architecture provides **defense-in-depth**: even if a sophisticated deepfake passes liveness detection, it must still match the enrolled voiceprint — and vice versa.

*(Kiến trúc hai lớp cung cấp phòng thủ theo chiều sâu: ngay cả khi deepfake vượt qua kiểm tra liveness, vẫn phải khớp voiceprint — và ngược lại.)*

---

## 📈 Roadmap

| Version | Status | Highlight |
| --- | --- | --- |
| **V1.2** | 🟢 **Current** | Commercial SOTA — EER 1.184%, Bank-grade verified |
| V1.5 (LMF) | 🔵 In progress | Large Margin Fine-tuning — targeting FRR@0.01% < 5% |
| V2.0 | 🟡 Planned | Next-generation architecture |

---

## 📞 Access & Licensing

This model is **Private** and available exclusively for enterprise partners under NDA.
*(Model nội bộ, chỉ cung cấp cho đối tác Doanh nghiệp ký NDA.)*

### Thương mại & Triển khai

- **License trọn gói** hoặc **qua API**
- Hỗ trợ tích hợp theo yêu cầu (triển khai, tối ưu hiệu năng, giám sát chất lượng)
- **Công ty Cổ phần SphinX** (sphinxjsc.com) được giao quyền **đóng gói, cung cấp API và phân phối**

### Bản quyền & License

**Thương mại / Proprietary**. Việc sử dụng, phân phối lại hoặc tạo bản phái sinh cần có chấp thuận bằng văn bản từ BrighTO Technology.

### Liên hệ

| Purpose | Contact |
| --- | --- |
| **Commercial Licensing** | `nguyen@brighto.ai`, `nghia@brighto.ai` |
| **API & Distribution** | `duc@sphinxjsc.com` (SphinX JSC) |
| **Technical Inquiries** | `nguyen@hatto.com` |

---

<div align="center">

### 🏆 BrightoSV Speaker Verification V1.2

**Commercial SOTA • Multilingual • Bank-Grade • Offline-Ready**

**EER 1.184%** · **20M Eval Pairs** · **600K Cohort** · **9+ Languages**

*Built in Vietnam 🇻🇳 • Engineered for the World 🌏*

</div>

---

> *This model card refers to BrightoSV Speaker Verification V1.2 (Commercial SOTA Release). All benchmark results are verified on internal evaluation sets comprising 20,000,000 scored pairs across 3,900+ speakers with strict bank-grade QA methodology and AS-NORM score normalization using a 600,000+ speaker cohort.*