Ctrlk
For the complete documentation index, see llms.txt. This page is also available as Markdown.

Changelog

[3.4.0]

  • FacephiActivationLoad is added to manage and activate the License.

[3.3.0]

  • Functions are added to improve License loading from cryptlex metadata.

  • The License registration process is improved during Docker initialization.

[3.2.1]

  • Minor License improvements.

[3.2.0]

  • New GET Endpoint /api/v1/config to obtain the service's current configuration.

  • New POST Endpoint /api/v1/config to update the service configuration without restarting it.

[3.0.1]

  • A License issue is fixed to ensure regulatory compliance and the correct operation of Voice Service.

[3.0.0]

  • The new verification algorithm is introduced, offering improved performance over previous versions. The previous algorithm remains available for compatibility reasons.

  • Accuracy of the liveness detection algorithm against voice cloning (voice clone liveness) is improved.

  • Accuracy of the liveness detection algorithm against repetition attacks (presentation/replay attack) is improved.

  • The error is fixed TOO_NOISY reported by the SNR Computer in files with long silences.

  • The OpenVINO backend is updated from version 2023.0 to 2024.4.

[2.9.0]

  • The Docker image version is updated to Ubuntu 24.04.

  • The number of threads in the service configuration is 1 by default.

  • A new verification algorithm with greater accuracy is added.

  • Accuracy of the verification and liveness detection algorithms is improved in noisy scenarios. The verification score may differ slightly from the previous version, but voice templates remain compatible.

  • The recommended threshold for the speech relative length quality check, in verification and liveness detection use cases, is now 0 (any relative speech length is allowed).

  • The SNR estimation algorithm is updated. The estimated values may differ from those in the previous product version, but the overall accuracy of the algorithm remains the same.

[2.8.0]

  • New property check_quality in the endpoints of enrollment and authentication to enable or disable quality checking.

  • New property check_liveness in the endpoint of authentication to enable or disable liveness checking.

[2.7.2]

  • Improvements to the service health check.

  • Improvements in License management.

[2.7.1]

  • Concurrency improvements in base64 decoding.

  • Some minor code improvements.

[2.7.0]

  • Some refactorings and performance improvements.

  • Concurrency improvements.

  • Improvements to the service health check.

  • New checks on License availability.

[2.6.0]

  • New Quality Check parameters:

    • minimum_speech_relative_length: minimum speech length required relative to the total audio duration.

    • maximum_multiple_speakers_detector_score: maximum number of simultaneous speakers.

  • New response fields with the values detected by the Quality Check:

    • snr_db_detected: detected signal-to-noise ratio (Signal-to-Noise Ratio).

    • speech_length_ms_detected: detected speech length.

    • speech_relative_length_detected: detected relative speech length.

    • multiple_speakers_score_detected: detected multiple speakers score.

[2.5.0]

  • Support for new audio formats: AAC, PCM ulaw and mulaw, WMA, MP4, FLAC, ALAC and AIFF.

  • Support for audio with a sampling frequency below 16 kHz.

  • Tracking events are integrated for Authentication.

[2.4.0]

  • Updated voice-sdk to v2.4.0.

  • Multithreading in the enrollment process.

  • Low-level exceptions are exposed to the user.

[2.3.1]

  • Updated voice-sdk to v2.3.1.

  • The exception and error messages are improved.

[2.3.0]

  • Updated voice-sdk to v2.3.0.

  • Updated facephi-rest-manager to v2.4.2.

  • Support for the AAC audio format is added.

  • New configuration parameter client_max_body_size to limit the maximum request body size.

[2.2.3]

  • Updated voice-sdk to v2.2.3.

  • The field is added operation_result to the Authentication and Enrollment result.

  • The field is added liveness_score to the Authentication and Enrollment result.

  • The field is added matching_score to the Enrollment result.

  • The field is added template_type to the Enrollment result.

  • The field is renamed probability from the Authentication result to matching_score.

  • The field is corrected template in the Enrollment result. The field template will be empty if the response is invalid.

  • The field is corrected template when multiple audio files were passed to Enrollment but the minimum number of audio files needed to generate a text-independent pattern was not reached.

[2.2.2]

  • Updated voice-sdk to v2.2.2.

  • Updated facephi-rest-manager to v2.4.0.

[2.2.0]

  • Support is added for voice-sdk v2.2.0.

  • The endpoints accept encrypted and unencrypted audio.

  • Quality Check can be configured in the request, with the new parameters minimum_snr_db and minimum_speech_duration_ms.

[2.1.0]

  • facephi-rest-manager is updated to version 2.1.3.

[2.0.0]

  • Support is added for voice-sdk v2.0.0, with encrypted audio and templates.

  • Refactoring of the REST API implementation.

  • New message format for Bad Request and Internal Server Error errors.

[1.3.1]

  • The HTTP status code is included in the JSON error response.

  • All endpoints are now lowercase.

  • Exception handling and messages are improved.

  • Exceptions are added for voice-sdk errors.

  • New endpoint api/v1/version.

  • New endpoint api/v1/health.

[1.3.0]

  • New License system.

PreviousDeployment on KubernetesNextRelease notes

Last updated