- Home
- AI Voice Biometrics Software
- Lifegence Voiceprint API
Lifegence Voiceprint API
What it is
Lifegence Voiceprint API is a free, MIT-licensed, self-hosted REST API for speaker verification and identification. Enrollment registers a speaker’s voiceprint. Verification compares submitted audio with one enrolled speaker for 1:1 authentication, while identification searches enrolled speakers for a 1:N match. SpeechBrain ECAPA-TDNN produces 192-dimensional voice embeddings. Supported audio formats are WAV, MP3, WebM and OGG, with recordings from 3 to 30 seconds and 5 to 10 seconds recommended. A Docker Compose quick start runs the API with PostgreSQL 16, pgvector and Redis 7. API keys are sent through the X-API-Key header. Enrollment includes consent fields, and a deletion endpoint removes associated voiceprint data. The security policy treats embeddings as sensitive biometric data and recommends encryption at rest, TLS 1.3 in transit and storing embeddings instead of raw audio. The current version has no liveness detection or anti-spoofing. The project warns that recognition is not guaranteed to be fully accurate and places legal compliance, including GDPR and BIPA, on users.
For and against
For
- Supports verification and identification
- Multiple common audio formats
- Docker Compose deployment
- Consent and deletion endpoints
- Open source under MIT License
Against
- No liveness or anti-spoofing measures
- Recognition is not guaranteed fully accurate
- Users handle GDPR and BIPA compliance
- Rate limits apply to API requests
The verdict
Lifegence Voiceprint API provides a self-hosted biometric API with enrollment, matching and deletion workflows. Its security guidance and lack of anti-spoofing require careful deployment decisions.
Who it suits
It suits developers deploying self-hosted voice authentication or speaker-identification services. Clients can include curl, Python, JavaScript and mobile apps.
On the record
- Accuracy limitation
- The project warns that voice recognition does not guarantee 100% accuracy.github.com · read 4 Oct 2026
- Audio input
- Supported audio formats are WAV, MP3, WebM, and OGG, with durations from 3 to 30 seconds.github.com · read 8 Oct 2026
- Authentication
- Verification compares submitted audio against an enrolled speaker and returns a score and confidence level.github.com · read 8 Oct 2026
- Biometric data
- The security policy says embeddings should be treated as sensitive biometric data and calls for encryption at rest and TLS 1.3 in transit.github.com · read 4 Oct 2026
- Client compatibility
- The architecture documentation lists curl, Python, JavaScript, and mobile apps as clients.github.com · read 8 Oct 2026
- Compliance note
- The README says compliance with applicable laws, including GDPR and BIPA, is the user's responsibility.github.com · read 8 Oct 2026
- Consent and deletion
- Enrollment requires consent information, and the API provides an endpoint to delete a speaker and associated voiceprint data.github.com · read 8 Oct 2026
- Deployment
- The project provides a Docker Compose quick start for running the API server.github.com · read 4 Oct 2026
- Enrollment
- The API enrolls speakers by registering a voiceprint from an audio file.github.com · read 8 Oct 2026
- Identification
- Identification searches across speakers and can return up to 20 results.github.com · read 8 Oct 2026
- License
- The repository identifies the project as MIT licensed.github.com · read 8 Oct 2026
- Purpose
- It is a self-hosted REST API for speaker verification (1:1) and identification (1:N).github.com · read 8 Oct 2026
- Rate limits
- The API specification lists limits of 10 enrollment, 60 verification, and 30 identification requests per minute.github.com · read 8 Oct 2026
- Security
- The security policy recommends TLS 1.3 in transit, encrypted database storage, API-key access control, and storing embeddings rather than raw audio.github.com · read 8 Oct 2026
- Security limitation
- The current version does not include liveness detection or anti-spoofing measures.github.com · read 8 Oct 2026
- Storage stack
- The documented stack includes PostgreSQL 16 with pgvector and Redis 7.github.com · read 4 Oct 2026
- Technology
- The project lists SpeechBrain ECAPA-TDNN with 192-dimensional embeddings, FastAPI, PostgreSQL with pgvector, Redis, and Docker Compose.github.com · read 8 Oct 2026
- Verification
- Its verification endpoint compares audio with an enrolled speaker for 1:1 authentication.github.com · read 4 Oct 2026
