Please don't forget to check the version history of LMS as it is a system that allows users to store speech recognition models.
Q1 2026 (Jan–Mar 2026)
Enhancements
- Flexible, Interval-Based Credit Consumption for SR Licensing
Usage-based SR licenses can now define a configurable "roundup interval" (for example, every 15 or 30 seconds of use equals one credit), giving customers more transparent and flexible consumption-based billing. Licenses without this setting continue to work exactly as before.
Improvements
- Separated, Configurable Audio and Text Logging for SR
Audio logs are now stored separately from text logs, with a configurable storage location for audio, making log management, storage planning, and log rotation easier — especially for on-premise deployments.
Bug Fixes
- Fixed Incorrect "No Match" Results with Uppercase Grammar Content
Resolved an issue where recognition incorrectly returned "no match" when grammar content was written in uppercase and an end-to-end model was in use.
Fixed Unnecessary Error Logs for License Attachment Data
Resolved an issue where recognition requests were writing unnecessary error logs related to license attachment data.
-
Fixed an Error in External SR Recognition Requests
Resolved an issue that could cause external SR recognition requests to fail. -
Fixed an Error When Requesting Model Version 0
Resolved an issue where external SR requests specifying model version 0 would return an error. -
Corrected Error Messaging for Non-Existent Model Versions
Fixed a mismatch between the error code and error message returned when an external SR request specifies a model version that doesn't exist.