Nijta
AI tool for voice anonymization, ensuring data privacy compliance.. [Contact for Pricing]
Last verified:
What is Nijta?
Nijta's Voice Harbor is an AI-powered voice data anonymization solution that removes biometric traces and personally identifiable information (PII) from audio recordings while preserving emotional tone, prosody, and broadcast-grade audio quality. Through its proprietary gen-AI speech-to-speech technology, the platform transforms voices to make speakers fully untraceable and unidentifiable, ensuring irreversible anonymization that complies with GDPR and the EU AI Act.
The tool is specifically built for newsrooms, media companies, contact centers, healthcare providers, defense organizations, and educational institutions that handle sensitive voice data. It enables journalists to protect confidential sources while publishing authentic-sounding audio, allows contact centers to use voice data for analytics without violating privacy, and helps healthcare AI companies process patient communications in HIPAA-compliant ways. The platform supports API integration for developers and plugins for Avid Pro Tools and Adobe software.
Key features include multilingual support (90+ languages for PII removal, English/French for biometric redaction), faster-than-real-time processing (Mini mode <0.5x RTF, Advanced mode ~0.75x RTF), 22 kHz high-fidelity broadcast-ready output, customizable voice characteristics (age, gender, accent), speaker diarization with the Monster system, and CNIL certification validating zero re-identification. The solution handles interviews, user-submitted audio, archival content, medical emergency calls, and customer service recordings at enterprise scale.
Voice Harbor has anonymized over 1 million recordings and offers both SaaS and on-premise deployment options. The platform eliminates hours of manual voice distortion work (pitch-shifting, filtering, re-recording) while maintaining clarity and emotion that traditional methods struggle to preserve.
Nijta pricing
Pricing model: Freemium
Purchase usage tokens to activate usage sessions. Download API or Plugin first, then buy tokens for processing. No free tier mentioned on website. Token-based pricing model for pay-per-use anonymization processing. Contact sales for enterprise pricing details.
Nijta pros
- Irreversible anonymization with zero re-identification guarantee
- CNIL-certified by French Data Protection Agency for legal compliance
- Preserves natural prosody, tone, and emotional expression
- Broadcast-ready 22 kHz high-fidelity audio output
- Faster-than-real-time processing (Mini mode <0.5x RTF)
- 90+ languages supported for PII removal
- API and plugin integration for Avid Pro Tools and Adobe
- Customizable voice characteristics (age, gender, accent)
- Handles overlapping speech with Monster diarization system
- Both SaaS and on-premise deployment options available
- SOC2 compliant for SaaS security
- Removes both biometric information and PII content
- Processes 1 hour of audio in 4 minutes with batch processing
- Supports unlimited number of speakers in diarization
- Over 1 million recordings already anonymized
- PII redaction error rate less than 1%
- Maximum 10 concurrent API requests per user
- Adapts to geographical and sociocultural conditions
Nijta cons
- Biometric redaction only available in English and French
- Does not work for children's voices accurately
- No real-time streaming audio support yet (planned Q3 2025)
- Age preservation feature not available currently
- Original emotion may degrade without fine-tuning
- Audio files limited to 500 MB maximum size
- No fine-tuning service for models on customer's site
- Small degradation in pronunciation sometimes observed
- Speech biomarkers for health attributes not certain to preserve
- Profane language filtering still in active research
Frequently asked questions about Nijta
What is speech anonymisation?
Speech anonymisation removes personal data from audio files according to global privacy regulations like GDPR. There are two types of personal data in audio: biometric information (non-verbal attributes like voice quality and timber that distinguish voices) and personally identifiable content (verbal cues like names, locations, organizations). Nijta's solutions handle both types to protect voice data privacy.
How many languages do you support?
The platform supports 90+ languages for removal of personally identifiable content and 2 languages (English and French) for removal of biometric information. New languages can be added in just one day by contacting the team.
What file formats do you support?
Nijta supports all popular audio and text formats. Users should refer to the documentation for a complete list of supported formats.
Does it work in real-time for streaming audio?
No, the solutions do not work in real-time currently. The real-time version is planned for release by Q3 2025.
Does the anonymisation work for children's voices?
No, the solution does not accurately work for children's voices. This is an active area of research at Nijta, and they are partnering with EdTech providers to build a robust solution for children's voices.
Could the age and gender of the speaker be preserved after anonymisation?
The gender of output voices can be controlled, but not the age. Nijta is working actively to provide the age preservation feature.
What is the processing time of your solution?
On a config with 15 cores, 45 GB RAM, and 1 x Tesla V100S, the system can anonymize 1 hour of audio in 4 minutes with 1 worker processing batches of 30 files (voice only). Mini mode runs at <0.5x real-time factor and Advanced mode at ~0.75x RTF.
What is the maximum size of audio files that could be sent to the API?
Audio files should not exceed 500 MB in size. The API will not accept files larger than this limit.
Does it work as a SaaS or on-premise?
Nijta provides both SaaS and on-premise solutions. For on-premise hosting, requirements include Linux (Ubuntu 20.04+), Intel Xeon Gold 6226R+ processor, 32 GB minimum RAM, 100 GB minimum disk space, and Nvidia Tesla V100S with 32 GB minimum GPU.
How many concurrent requests could the API process?
The system supports concurrent requests, allowing a maximum of 10 simultaneous requests per user without degrading processing time.