Changelog
[v1.3.2] - 2026-09-17
- Aligned the public APIs and integration methods on iOS and Android. New integrations should use the APIs recommended in the current documentation.
- Added authentication-mode selection APIs supporting
default,online,offline, andall; iOS and Android expose the corresponding platform-specificverifyAuthoverloads. - Fixed an iOS bug where the offline-model readiness check could return an incorrect result after an app downloaded the offline model package outside the SDK; the corresponding unit tests passed.
- Upgraded the offline SDK to
1.2.2. The offline model fixes four issues: ASR occasionally returned an empty final result;<>symbols were removed from recognized text; ASR results containing only punctuation no longer generated bubbles; and the Japanese sentence-segmentation 1 self-test issue was fixed.
[v1.3.1] - 2026-09-11
- Added
durationandoffsetfields to online ASR results to identify the speech-segment duration and its position in the audio stream; use thebubbleEndevent to confirm that a bubble has ended. - Added
partialandstableresult-delivery modes for offline translation so applications can choose the experience that fits their use case. - Added runtime source-language and target-language switching for offline translation without destroying and rebuilding the entire translation pipeline.
- Added the
offline_bubble_endevent for accurately marking the end of an offline-translation bubble. - Added runtime switching among the
recognize,toText, andtoSpeechcapability tiers for offline translation; the required offline models must be downloaded before upgrading a tier. - Added a low-latency audio mode for online one-to-one translation, including PCM input by left and right channel.
- Added access to translated PCM audio so applications can control playback, audio processing, and routing; audio callbacks can be configured to run on a background thread.
- Added an online-translation sensitive-word redaction switch, enabled by default; confirm the redaction policy before upgrading if the business needs to display or process the original text.
- Added Essential, Diagnostic, and Trace diagnostic-log levels, with encrypted persistence and session-level log isolation.
- Optimized final-translation assembly for long offline speech, TTS callback backlog handling, and bubble ordering.
- Improved online-translation RTM reconnection, parsing, and language switching during reconnection.
- Optimized offline-translation performance and fixed known issues.
- Aligned the language-code settings for
sourceLanguageandtargetLanguagewhen creating a room on iOS and Android.
[v1.2.8] - 2026-08-21
- Fixed Android release-package obfuscation errors to ensure the compiled code runs correctly.
- Fixed abnormal Android timeout responses to prevent no-response exceptions.
- Updated Android native dependencies for 16 KB page-size support, addressing compatibility issues and potential crashes on some devices.
[v1.2.3] - 2026-08-03
- Android diagnostic logs (including RTC-related logs) are now stored in the
files/directory within the app sandbox internal storage.
[v1.2.1] - 2026-07-29
- Refined SDK error codes: the original error code system has been split into more granular codes, letting developers locate and handle different abnormal scenarios more precisely for faster troubleshooting and better user experience.
- Support setting the translation engine when creating a room: you can now specify the end-to-end or three-stage (ASR → MT/LLM → TTS) recognition/translation engine when creating a room, so you can pick the translation pipeline that fits your business scenario; see Android:
TmkTranslationRoomConfig.onlineRecognizeEngineand iOS:TmkTranslationRoomConfig.recognizeEngine.
[v1.2.0] - 2026-07-21
As the first official release, Tmk Translation SDK provides unified real-time speech translation capabilities for iOS and Android, helping developers quickly build multilingual real-time speech translation apps.
Key features
- Online translation: 55 languages, 109 accents.
- Offline translation: 11 major languages, fully functional without a network.
- Integrated speech recognition (ASR), machine/large-model translation (MT/LLM), and speech synthesis (TTS).
- Unified API with online/offline mode switching, state management, error handling, offline model management, and problem diagnostics.
Release contents
- Tmk Translation SDK v1.2.0 (iOS / Android, integrated via CocoaPods / Maven Central)
- SDK docs: /docs/v1.2.0/intro
- Sample Demo: https://github.com/timekettle/tmk-translation-sdk-samples