Skip to main content

Mobile speech SDK selection and download

Download NUI mobile SDK packages for Model Studio speech services. Select a package by platform and application ID, and review compatibility requirements and release notes.

Select a package

The version number, such as 2.7.4, identifies an SDK release. The application ID, such as 037 or 047, identifies a package feature set.
V2.7.3 and earlier do not support the new Model Studio workspace domains. Using these domains causes SDK initialization or connection failures. Use the following legacy endpoints:
  • China (Beijing): wss://dashscope.aliyuncs.com/api-ws/v1/inference
  • Singapore: wss://dashscope-intl.aliyuncs.com/api-ws/v1/inference
V2.7.4 and later support the new workspace domains.
The latest standard packages with the following application IDs support these Model Studio cloud capabilities:
  • Qwen-Audio-TTS/CosyVoice speech synthesis
  • Sambert speech synthesis
  • Qwen-Audio-ASR/Fun-ASR speech recognition / recorded speech recognition
  • Gummy speech recognition / translation
  • Qwen-Livetranslate-Realtime audio translation
  • Qwen-Audio-Realtime conversations
On-device capabilities vary by application ID as follows.
Application IDOn-device speech synthesisOn-device wake word detectionOn-device acoustic echo cancellation and voice activity detection
039SupportedSupportedNot supported
038SupportedNot supportedNot supported
037Not supportedNot supportedNot supported
049SupportedSupportedSupported
047Not supportedNot supportedSupported

Download SDKs

Select an SDK for your platform and required features.
Use a standard package for regular integration. If your project already integrates an SDK with the same name, use a renamed package to resolve naming conflicts.
  • Android
  • iOS
  • HarmonyOS
  • Flutter
Platform update date: 2026-10-27.Package name: com.alibaba.idst.nui. Requires Android 5.0 or later (API Level 21). Supported architectures: armeabi-v7a, arm64-v8a, and x86_64.
  • Standard packages
  • Renamed packages
Application IDVersionDownloadMD5
0392.7.6V2.7.6-039-20261027_Android_TLS.tar.gze4267e1e2e159a58bf0253abe4c057da
0382.7.6V2.7.6-038-20261027_Android_TLS.tar.gzf1cfea702d60dec1dd9d704386643838
0372.7.6V2.7.6-037-20261027_Android_TLS.tar.gz4aa0469122d0656d6ffea299d52e0ff7
0492.7.6V2.7.6-049-20261027_Android_TLS.tar.gz31490f3e3cf230ec14a3c1ae5b01a08c
0472.7.6V2.7.6-047-20261027_Android_TLS.tar.gzc1191f473ca119a32a8f5078b9b4fc28

Release notes

V2.7.6

  • Added support for new Qwen-Audio-ASR models for non-real-time speech recognition.

V2.7.5

  • Improved the ability to resume real-time speech recognition after a network interruption.
  • Added support for new Qwen-Audio-ASR models.
  • Fixed an issue that prevented the generic parameter-setting API from taking effect.

V2.7.4

  • Added Qwen-Audio-Realtime conversations.
  • Added support for new Model Studio workspace domains.

V2.7.3

  • Added real-time audio translation.
  • Added resumption after network disconnection and disconnection error reporting.

V2.7.2

  • Fixed bugs.

V2.7.1

  • Fixed bugs.

V2.7.0

  • Added support for Model Studio speech recognition and synthesis.
  • Added decoding of Opus and MP3 to PCM for speech synthesis.
  • Reduced the OpenSSL footprint to decrease package size.