Release notes
This topic provides the release notes for Real-Time Communication.
March 2026
Category | Feature | Description | Platform | Release date |
server-side | Automatic hang-up for unresponsive calls | You can now configure an automatic hang-up policy for AI-powered outbound calls. This policy disconnects unresponsive calls after a specified duration to improve call efficiency. | server-side | March 31, 2026 |
Additional PSTN call information | The agent instance details page now displays configuration information for PSTN outbound and inbound calls. | server-side | March 31, 2026 | |
Dynamically update Model Studio parameters during calls | You can now dynamically modify Model Studio large model parameters during digital human calls and video calls. | server-side | March 31, 2026 | |
Video content moderation with side-channel models | The visual understanding feature now uses a side-channel model for video content moderation. | server-side | March 31, 2026 | |
Voice cloning support for TTS nodes | TTS nodes now support voice cloning with CV3, CV3.5, and Qwen3 models, enabling personalized speech synthesis. | server-side, console | March 31, 2026 |
January 2026
Feature category | Parameter | Description | Supported client | Release date |
server-side | Voiceprint Denoising 2.0 | Uses next-generation voiceprint recognition to isolate the speaker's voice from background noise, significantly improving call clarity. Supports both pre-registration and seamless registration modes. | server-side, Console | 2026-01-29 |
Multi-Model Speech Recognition | Provides four built-in STT speech recognition models optimized for different scenarios, improving recognition accuracy. | server-side, Console | 2026-01-29 | |
Model Studio Large Model Integration | Integrates the Fun and Qwen3 models from Model Studio to expand voice interaction capabilities. | server-side, Console | 2026-01-29 | |
Multi-Model Speech Synthesis | Provides four built-in TTS speech synthesis models for more natural and diverse voice output. | server-side, Console | 2026-01-29 | |
Streaming Text-to-Speech | Generates audio as text is input, which reduces latency and enhances real-time performance. | server-side, Console | 2026-01-29 | |
English Replacement Word Optimization | TTS speech synthesis supports configurable English replacement words for multilingual expressions. | server-side | 2026-01-29 | |
Call Quality Data Visualization | Data Compass provides per-call quality analysis, allowing users to evaluate the performance of individual calls. | Console | 2026-01-29 | |
Smart Backchanneling | Generates intelligent, contextual backchanneling responses like "uh-huh" and "okay" to make conversations more natural. | server-side | 2026-01-29 | |
Audio and Video Call | Mobile Client Experience Optimization | Improves the mobile client demo for a better onboarding experience and smoother interactions. | Official Agent | 2026-01-29 |
December 2025
Feature category | Feature name | Description | Supported client | Release date |
server-side | Intelligent sentence segmentation optimization | Optimizes the semantic sentence segmentation logic to reduce wait times and improve conversational flow. | server-side | 2025-12-31 |
Conversation timestamp optimization | Conversation logs now include a timestamp for when the agent begins to respond, simplifying tracing and analysis. | server-side | 2025-12-31 | |
Context and sentence segmentation optimization | Optimizes context logic and sentence segmentation tags to improve conversational coherence and naturalness. | server-side | 2025-12-31 | |
Large model invocation statistics | Exposes the UID for large model invocations to support tracking and analysis of token usage. | server-side | 2025-12-31 | |
Audio/video call | Linux SDK support | Provides an SDK and sample code for the Linux platform to enable client-side deployment on Linux. | Linux | 2025-12-31 |
HarmonyOS AI Call support | Introduces the AI Call SDK for HarmonyOS for native application development. | HarmonyOS | 2025-12-31 |
November 2025
Category | Feature | Description | Platform | Date |
server-side | ChatGPT integration | The LLM node now integrates with ChatGPT to expand conversational capabilities. | server-side, console | November 18, 2025 |
Agent call recording | Enables recording agent calls for later review and analysis. | server-side, console | November 2025 | |
US West 1 (Silicon Valley) region launch | The service is now available in the US West 1 (Silicon Valley) region, reducing latency and improving stability for international users. | server-side, console | November 27, 2025 | |
Outbound call transfer to human agent | Enables transferring outbound calls to a human agent based on predefined rules, increasing service flexibility. | server-side | November 27, 2025 | |
No-response reminder | Automatically sends a reminder when a user is unresponsive for an extended period to prevent conversation interruptions. | server-side | November 27, 2025 | |
AI response timeout announcement | Automatically plays a transitional message when AI processing times out to enhance the user experience. | server-side | November 27, 2025 | |
Smart hang-up | Enables automatically ending calls based on response content, suitable for specific business scenarios. | server-side | November 18, 2025 | |
Hotword recognition | FireRed ASR now supports hotword configuration to improve keyword recognition accuracy. | server-side | November 18, 2025 | |
audio and video call | New web-based demo | The console now features a web-based demo for quick and easy trials. | console | November 27, 2025 |
AICall enhancements for Linux | Improves the stability and compatibility of the AICall feature on the Linux platform. | Linux | November 18, 2025 |
October 2025
Feature category | Feature name | Description | Supported platform | Release date |
server-side | End-to-end model support | Leverages end-to-end speech models to streamline the processing workflow and reduce latency. | server-side, console | 2025-10-30 |
Sensitive word handling optimization | Optimizes sensitive word detection and handling to enhance compliance and security. | server-side | 2025-10-30 | |
Interruption word logic optimization | Optimizes interruption word recognition to improve conversational responsiveness. | server-side | 2025-10-30 | |
PSTN inbound callback support | Allows agents to trigger callbacks on inbound PSTN calls, simplifying integration with business systems. | server-side, console | 2025-10-30 | |
PSTN data filtering and analysis | Data Compass now supports filtering by PSTN data for independent analysis of traditional phone line performance. | console, Data & Billing | 2025-10-30 | |
ChatTTS performance improvements | Optimizes speech synthesis performance to reduce latency and improve real-time responsiveness. | server-side | 2025-10-30 | |
VAD optimization | Further optimizes the VAD11 voice activity detection model to better distinguish between speech and silence. | server-side | 2025-10-30 | |
audio and video call | Demo UI style upgrade | Updates the visual style of the demo for an enhanced user experience and a more professional appearance. | Native & Web | 2025-10-30 |
September 2025
Category | Feature | Description | Platform | Date |
server-side | Instance ID in call messages | Inbound and outbound call messages now include the instance ID to simplify managing and tracking multiple instances. | server-side, RTC | September 2025 |
PSTN latency optimization | Reduced PSTN line transmission latency from 5–6 seconds. | server-side, RTC | September 23, 2025 | |
New conversational voices | You can now preview and select four new natural, conversational voices in the console. | console | September 23, 2025 | |
API key security enhancement | To improve console security, API keys are no longer displayed in plaintext. | console | September 23, 2025 | |
audio/video call | Smoother digital human animations | Fixed abrupt motion changes in digital human animations to make them smoother and more natural. | server-side | September 23, 2025 |
message conversation | Disconnection handling improvement | Improved the reconnection experience for disconnected message conversations. | Native & Web | September 23, 2025 |
August 2025
Category | Feature | Description | Supported platform | Release date | Related documents |
server-side | third-party TTS optimization | Lets you set delimiters and schedule pushes. | server-side | 2025-08-14 | |
custom sensitive words | Lets you define your own list of sensitive words. | console | 2025-08-14 | ||
Inbound call welcome message | Lets you use an audio file as the welcome message. | server-side | 2025-08-14 |
July 2025
Feature category | Feature name | Description | Supported platform | Release date | Related documents |
server-side | MiniMax Turbo2 Model Support | Adds support for the MiniMax Turbo2 model. | console | 2025-07-17 | |
Voiceprint recognition 2.0 | The voiceprint recognition 2.0 system enables accurate speaker identification. | server-side | 2025-07-31 | ||
Data Compass | Data Compass is now in public beta. | console | 2025-07-31 | ||
Inbound call | Lets users call an AI agent by dialing a phone number. | console | 2025-07-31 | ||
Lingjing Digital Human | Adds support for cloud-rendered digital humans. | console | 2025-07-31 | ||
audio and video call | RTOS SDK | Designed for AI-powered smart hardware, this SDK enables an ultra-low latency, full-duplex conversational experience. | 2025-07-31 |
June 2025
Category | Feature name | Description | Platform | Release date | Related documentation |
Server-side | Dynamic parameters for STT | The in-house STT node now accepts dynamic parameters from the client-side. | Server-side | 2025-06-17 | |
Dynamic parameters for Model Studio | You can now set the prompt for Model Studio agents at runtime. | Server-side | 2025-06-17 | ||
Agent callback | Adds support for callbacks on outbound calls. | Console | 2025-06-30 | ||
Semantic segmentation | User input can now be semantically analyzed to understand its structure, intent, and semantic boundaries. | Server-side | 2025-06-30 | ||
Audio/video calls | Agent broadcasting | Enables agents to execute commands in sync with their audio broadcasts. | Server-side | 2025-06-17 | |
Outbound call experience | Outbound calls now include a daily quota of 20 trial calls. | Server-side | 2025-06-30 |
May 2025
Feature category | Feature name | Description | Platform | Release date | Related documents |
server-side | agent callback | Adds support for sentence-by-sentence audio callbacks for agents. | server-side | 2025-05-16 | |
interruption optimization | Optimizes single-word interruption for agents. | server-side | 2025-05-16 | ||
speech recognition hotword | Adds support for instance-level hotword configuration for agents. | server-side | 2025-05-16 | ||
offline video analysis | Adds support for re-analyzing offline videos. | server-side | 2025-05-31 | ||
memory turns | Workflows support configuring memory turns. | Console | 2025-05-31 | ||
message conversation | rich message input and output | Adds support for outputting code blocks and tables in Markdown format. | Web | 2025-05-31 |
April 2025
Feature category | Feature name | Description | Platform | Release date | Related documents |
server-side | ASR silence timeout | Lets you adjust the ASR silence timeout at runtime. | server-side | 2025-04-16 | |
Official agent optimization | Improves latency and prompt settings for the official agent. | server-side | 2025-04-16 | ||
audio and video call | Web demo for audio and video call | The Web demo for audio and video call is now available on the international site. | Web | 2025-04-28 | |
Support for emotional digital humans | Lets you customize digital humans with multiple emotional expressions. | Native | 2025-04-28 | ||
Barge-in prevention | Lets you prevent barge-in for specific audio and video content. | Native & Web | 2025-04-28 |
March 2025
Feature category | Feature name | Description | Platform | Release date | Documentation |
messaging | Rich message input and output | Supports image input and the output of rich content, including images, hyperlinks, and more. | Native & Web | March 31, 2025 | |
audio and video calls | Experience optimization | Reduces the call connection time to under 1 second. | Native & Web | March 31, 2025 | |
international site | Launches the Singapore site. | Native | March 13, 2025 | ||
console | Agent Callback | Lets you configure sentence-by-sentence callbacks for an agent. | console | March 31, 2025 | |
STT node upgrade | The STT node now supports custom integrations. | console | March 31, 2025 | ||
TTS node upgrade |
| console | March 13, 2025 | ||
Multimodal large language model integration with messaging | Supports integrating a multimodal large language model from the console. | console | March 13, 2025 |
February 2025
AICallKit SDK v2.1.0 introduces the following features:
Feature category | Feature name | Description | Platform | Release date | Related documentation |
chat | Chat agent | Adds support for the chat agent on the Web platform. | Web | 2025-02-28 | |
Chat history optimization | The chat history now synchronizes with audio and video calls. | Native | 2025-02-28 | ||
Support for DeepSeek | The chat agent now supports DeepSeek and returns its reasoning process. | Native | 2025-02-28 | ||
Business parameter passthrough | You can now pass business parameters to downstream business systems. | Native | 2025-02-28 | Pass user information to Model Studio or self-developed models | |
Proactive message sending | You can now proactively send text or custom messages to the client. | Native | 2025-02-28 | ||
audio and video call | Keyword-based interruption | You can now set specific keywords to interrupt the agent's speech. | Native & Web | 2025-02-28 | |
Large model prompt | You can now configure large model prompts at the instance level. | Native & Web | 2025-02-28 | ||
Quality inspection mode | In visual understanding calls, you can now specify the capture time and the number of frames for quality inspection. | Native & Web | 2025-02-20 | ||
Server-side message callback | You can now send messages to your registered callback service. | Native & Web | 2025-02-20 | ||
console | LLM support | The LLM node in a workflow now supports configuring Tongyi Qwen-Stardust. | Console | 2025-02-20 |
January 2025
Feature name | Description | Supported platform | Release date | Related documentation |
Message conversation | Adds a message conversation call type to interact with agents in a chat box. | Native | 2025-01-23 | |
LLM input passthrough | Adds an API to pass input content to the LLM. | API | 2025-01-23 | |
Voice configuration for third-party TTS plugins | The TTS third-party plugin (MiniMax) now supports voice configuration. | Console | 2025-01-23 | |
Language configuration for third-party STT plugins | The STT third-party plugin (iFlytek) now supports language configuration. | Console | 2025-01-23 | |
Noise reduction optimization | Fixes an issue where noise reduction removes the first few words. | Native and Web | 2025-01-23 | |
Agent/workflow console page optimization | The agent and workflow pages in the console now display resources from all regions by default. | Console | 2025-01-23 |
December 2024
Feature | Description | Platform | Release date | Documentation |
Emotion recognition | Agents can now recognize user emotions and deliver emotionally-aware responses. | Native & Web | 2024-12-31 | How to implement emotion recognition and generation in an agent |
One-click reporting enhancement | The one-click reporting feature is now more prominent in the demo. | Native & Web | 2024-12-31 | |
STT node configuration enhancement | You can now configure the language and silence duration in the STT node. | Console | 2024-12-31 | |
TTS node configuration enhancement | You can now adjust the speech rate and pitch for the default voice in the TTS node and preview the changes. | Console | 2024-12-31 | |
Agent creation enhancement |
| Console | 2024-12-31 |
November 2024
Feature | Description | Platform | Release date | Documentation |
Web demo | The web demo is now available on mobile devices. | Web | 2024-11-15 | |
Custom agent avatar | You can now upload an image to use as your agent's avatar. This avatar appears during voice calls. | Console | 2024-11-15 | |
Alibaba Cloud Model Studio integration | You can now use AI real-time interaction applications published on Alibaba Cloud Model Studio. | Console | 2024-11-30 | Alibaba Cloud Model Studio ✖ AI real-time interaction best practices |
Manual takeover | When an agent struggles or a situation requires a critical decision, a human agent can take over the conversation. | API | 2024-11-30 | |
Voice cloning | You can now use CosyVoice for voice cloning. | API | 2024-11-30 |
October 2024
Feature name | Description | Supported platform | Release date | Related documentation |
visual understanding call | Lets users interact directly with an intelligent agent through video. The agent analyzes the video content to identify and interpret what it sees. | Native & Web | 2024-10-31 | |
walkie-talkie mode | Users can enable walkie-talkie mode at startup or during a call. In this mode, they press a button to speak to the intelligent agent. | Native & Web | 2024-10-31 | |
voiceprint noise reduction | In multi-speaker conversations, the intelligent agent uses the main speaker's voiceprint characteristics to better capture their voice and reduce background noise. | Native | 2024-10-31 | |
STT/TTS integration with Alibaba Cloud Model Studio | The STT/TTS node now supports large models from Alibaba Cloud Model Studio. | console | 2024-10-31 | |
real-time captions enhancement | Improves the display of real-time captions when smart sentence segmentation is enabled. | Native & Web | 2024-10-31 |
September 2024
Feature name | Description | Platform | Date | Documentation |
Digital human call optimization | Improves digital human calls by optimizing concurrent channels, avatars, and real-time subtitles. | Native | 2024-09-15 | |
Intelligent sentence segmentation | Agents now automatically segment long or complex sentences, improving readability and the user experience. | Native & Web | 2024-09-27 | |
SMS verification | The demo now supports SMS verification for login. | Native & Web | 2024-09-27 | |
Digital human call | The web demo now supports digital human calls. | Native & Web | 2024-09-27 | |
Usage statistics | The console now displays usage statistics for the AI real-time interactive service. | Console | 2024-09-27 | |
AICallKit SDK | The AICallKit SDK enables rapid development of AI real-time interactive capabilities. | Native & Web | 2024-09-27 |
August 2024
Parameter | Description | Platform | Release date | Related documentation |
third-party TTS plugin | The TTS workflow node now supports integration with the MiniMax third-party plugin. | console | 2024-08-16 | |
third-party STT plugin | The STT workflow node now supports integration with the iFlytek third-party plugin. | console | 2024-08-16 | |
graceful stop | Allows an agent task to complete its current output before terminating, preventing abrupt interruptions in the conversation. | Native & API | 2024-08-16 | |
server-side callback | You can now receive real-time callbacks about the agent's operational status. | console | 2024-08-16 | |
web demo | The web demo for AI real-time interaction is now available. | Web | 2024-08-31 | |
digital human call | AI real-time interaction now supports the digital human call type. | Native | 2024-08-31 | |
ASR hotwords | You can now define business-specific hotwords to improve your AI agent's speech recognition accuracy. | console | 2024-08-31 | |
AICallKit SDK | You can now integrate the AICallKit SDK to quickly implement AI real-time interaction. | Native | 2024-08-31 | |
AppServer | You can now use the AppServer source code to implement the server side of AI real-time interaction. | Native & Web | 2024-08-31 |
July 2024
Feature name | Description | Supported platform | Release date | Related documentation |
Real-time call | Enables reliable, low-latency global calls with AI agents, powered by Alibaba Cloud's real-time communication solution. | Native | July 31, 2024 | |
Real-time workflow | Provides a visual builder to flexibly orchestrate the AI agent's workflow.
| Console | July 31, 2024 | |
Intelligent interruption | The AI agent intelligently detects when a user intends to interrupt the conversation. | Native | July 31, 2024 | |
Intelligent noise cancellation | The AI agent automatically filters ambient noise from the user's audio input. When multiple people speak simultaneously, the agent prioritizes the loudest voice. | Native | July 31, 2024 | |
AI agent status | Provides real-time status updates for the AI agent:
| Native | July 31, 2024 | |
Real-time captions | Displays a real-time text transcription of the conversation between the user and the AI agent on the client. | Native | July 31, 2024 | |
Welcome message | Lets you set a welcome message that plays at the start of a conversation with an AI agent. | Console and API | July 31, 2024 | |
Data archiving | Lets you archive call recordings between users and AI agents in segments to Object Storage Service (OSS). | API | July 31, 2024 |