Skip to content

Offline ASR SDK Version Overview

CI13XX_SDK_ASR_Offline_V2.2.0

Version 2.2.0 โ˜žMigration Guide

What’s new:

  1. Support switching acoustic models. For example, when supporting both Chinese and English, switching between different versions of DNN pro5 models is supported. Switching between pro5 and pro is not supported.
  2. Added flash protection.
  3. Removed some algorithm macros that are no longer supported in offline versions. Updated tool to V3.9.9. You can query firmware checksums and view partition addresses.

CI13XX_SDK_ASR_Offline_V2.1.14

Version 2.1.14 โ˜žMigration Guide

What’s new:

  1. Support various pro models, improving adaptability and recognition stability.
  2. Fixed incomplete playback combinations under specific conditions.
  3. Default synthesized prompt uses -b16 to reduce voice asset size.
  4. Recognition speed: ADAPTIVE_CNT_ENABLE 1 = adaptive cnt; 0 = faster response, but you must set reasonable cnt values in cmd_info entries. Packaging tool updated to V3.9.7.

CI13XX_SDK_ASR_Offline_V2.0.10

Version 2.0.10 โ˜žMigration Guide

What’s new:

  1. Initial release of Natural Speech SDK.
  2. Optimized user_config.h for beginners.
  3. This SDK can be used for offline Natural Speech solutions and command-word solutions.
  4. Pro models offer higher accuracy and better noise robustness; prioritize the highest version number when choosing models.
  5. This SDK retains only the project offline_asr_pro_sample; other samples (e.g., command self-learning) will be added in later versions.
  6. Updated packaging tool.

CI13XX_SDK_ASR_Offline_V1.12.16

Version 1.12.16 โ˜žMigration Guide

What’s new:

  1. Improved recognition stability to avoid extremely rare cases of a specific word performing poorly.
  2. Broader baud rate compatibility.
  3. Optimized code logic when self-learning and AEC coexist. Important: This is the final version of the fixed command-word SDK; no new features will be added afterward.

CI13XX_SDK_ASR_Offline_V1.11.7

Version 1.11.7

What’s new:

  1. Updated recognition library and optimized memory usage.
  2. When volume is set to 0, the DAC digital gain is also set to 0 to eliminate faint sounds.
  3. Fixed an issue where UART data might not finish sending before reset under specific conditions.
  4. Packaging/upgrade tool updated to V3.8.3.
  5. Fixed an issue where firmware info was not immediately refreshed when jumping directly to the upgrade interface after packaging.
  6. Firmware version number in the packaged upgrade tool now adds dots to the filename only; it is not written into the partition table.