Recording Summary Agent tutorial

​

Prerequisites

• You have created a multimodal application in Model Studio.

​

Add the Recording Summary Agent

In the Multimodal Interaction Console, choose AgentRecommended Applications, select Recording Summary, and click Confirm to add it.

​

Activate the Tingwu service

When you use the Recording Summary Agent for the first time, you must activate the Tingwu service. Go to Tingwu - Intelligent Summary to activate and publish the application.

1. In the Tingwu - Intelligent Summary application, click Activate Now in the upper-right corner.
2. On the activation page, select Tingwu Agent Pay-As-You-Go and click Buy Now to complete the activation.
3. On the order confirmation page, confirm the Tingwu Agent Service order information (billing method: Pay-As-You-Go). Read and agree to the Product Service Agreement and Service Level Agreement, then click Activate Now to complete the activation.
4. Click Create Application in the upper-right corner of the Tingwu - Intelligent Summary console to configure and publish the application.
5. On the Debug Configuration tab, complete the following settings:
   1. In Select Audio Input Method, select the desired mode (for example, Real-time Transcription).
   2. Select the Transcription Model (for example, paraformer-v2).
   3. In Transcription Configuration, set the number of Speaker Separation speakers and Recognition Language.
   4. In the Translation area, set the Content Translation target language.
6. Click Publish in the upper-right corner to publish the application.

​

Configure the application

After you create and publish an application in the Tingwu - Intelligent Summary console, select the corresponding Tingwu application in the Multimodal Interaction Console. To adjust application capabilities, configure the settings in the Tingwu - Intelligent Summary console.

​

Real-time recording summary settings

In the right-side configuration panel, you can set the start command, exit/pause wake word, and resume command for real-time recording summary. Exit/Pause wake word — During recording, the system recognizes exit or pause commands only after detecting the wake word, which prevents accidental exits. The default wake word is Xiao Yun. (Recommended) Add the wake word to Tingwu's hot words to improve recognition accuracy. Each command supports clicking + to add multiple entries, with a character limit of 100. The following are representative command examples for each action. Various natural language phrasings are supported:

• Start command — "Start real-time transcription", "Open real-time transcription to record this conversation", "Enable real-time transcription to record group discussion"
• Pause command (custom wake word) — "Xiao Yun, pause transcription", "Xiao Yun, pause real-time transcription", "Xiao Yun, pause for now"
• Resume command — "Resume recording", "Continue real-time transcription", "Continue transcription"
• Exit command (custom wake word) — "Xiao Yun, stop real-time transcription", "Xiao Yun, end real-time transcription", "Xiao Yun, stop transcription and summarize this discussion"

​

Recording file summary settings

You can set the start command, exit command, pause command, and resume command for recording file summary. The following are representative command examples:

• Start command — "Start recording and summarize", "Record and summarize this meeting", "Open recording and summarize the current conversation"
• Exit command — "Stop recording", "End recording", "End recording, help me summarize and translate to English"
• Pause command — "Pause recording", "Wait before recording", "Pause for now"
• Resume command — "Resume recording", "Continue recording", "You can continue recording"

​

Conversation experience

After you complete all configurations, you can experience Tingwu's recording and meeting summary capabilities in the Multimodal Interaction Console.

​

Start real-time transcription

Use a configured start command to begin transcription. For example, say "Enable real-time transcription". You can also view bilingual subtitles. After you send the command, the system displays meeting_state_change command sent successfully. The AI confirms that real-time speech-to-text has been enabled and that a recording summary will be automatically generated after transcription ends. A real-time transcription area appears at the bottom of the voice interaction experience panel on the right, displaying transcription timing and Single/Complete toggle buttons.

​

Exit real-time transcription

Use the configured wake word combined with the exit command to exit real-time transcription. For example, say "Xiao Yun, exit real-time transcription". Click View Analysis Results to go to Tingwu and view the intelligent summary. The Tingwu Intelligent Summary page displays the speech-to-text transcription and translation results on the left, with playback controls at the bottom including the play button, progress bar, and speed settings. The right side shows the analysis results area with three tabs: Quick Overview, Mind Map, and Custom Prompt. You can view automatically extracted keyword tags, full-text summaries, chapter overviews, and key point reviews.