Manage projects

Prerequisites

Intelligent Speech Interaction uses a pay-as-you-go billing model, where you are charged based on the actual duration of audio processed, measured down to the second. You do not need to purchase fixed-specification instances in advance.

Activate the service by following the instructions in Step 3: Activate the service. You can also navigate to Service Management and Opening in the left navigation pane of the Intelligent Speech Interaction console to manage service activation. After the service is activated, you can create projects and obtain Appkeys.

Create a project

1. Sign in to the Intelligent Speech Interaction console.

2. In the left navigation pane, click All Projects.

3. On the My All Projects page, click Create Project.

4. In the Create Project dialog box, enter a Project Name, select a Project Type, and optionally add a Project Scenario Description.

   Available project types: speech recognition + text-to-speech + speech analysis, speech recognition only, text-to-speech only, and on-device solution.

   The following table describes how each project type affects the console display:

   Project type	Console display
   speech recognition + text-to-speech + speech analysis	Displays both the Speech Recognition and speech synthesis service configuration areas. This is the default type.
   speech recognition only	Displays only the Speech Recognition configuration area.
   text-to-speech only	Displays only the speech synthesis service configuration area.
   on-device solution	Displays the SDK quota management area.

   The project type only determines which configuration areas appear on the project function configuration page. It does not restrict API calling capabilities. For example, a project of type speech recognition only can still call speech synthesis APIs using the same Appkey. When you select a speech-recognition-related type, all speech recognition APIs — including recording file recognition, real-time speech recognition, and one-sentence recognition — are available by default. You can call multiple speech recognition APIs under the same project without creating separate projects for each.

   All speech recognition APIs — including recording file recognition, real-time speech recognition, and one-sentence recognition — are available by default under each project. You can call both speech recognition and speech synthesis APIs using the same Appkey without creating separate projects for each service.

5. After the project is created, find the project and its appkey on the My All Projects page.

Configure a project

Speech recognition

These steps apply when Project Type is set to speech recognition only or speech recognition + text-to-speech + speech analysis.

1. Click Project Function Configuration to the right of the target project.

2. In the speech recognition (ASR) section, select a base model or a self-learning model.

   • Click Modify Configuration, select a base model for your use case, test it online, and click Confirm to Use.

     On the model selection page, models are organized into the following tabs:

     Tab	Description
     Except Telphone	Models for non-telephone scenarios with a sampling rate of 16 kHz.
     Telphone	Models for telephone scenarios with a sampling rate of 8 kHz.
     Latest Releases	Recently published models.
     Most Frequently Used	Popular models based on usage frequency.

     Under the Except Telphone and Telphone tabs, models are further categorized by industry, such as General, Education, Judiciary, Medical, Conference & Speeches, News & Media, Entertainment & Video, Smart Home, Social, Automotive, and Others. To use a specific dialect model (such as Cantonese) or a scenario-specific model (such as a telephone model), browse the categories and select the appropriate model.

     Important

     The speech recognition service does not support switching models dynamically through API parameters. Each project is bound to a single speech recognition model. To use different models, create separate projects with different model configurations, and use the corresponding Appkey for each project when calling the API.

     No single model supports mixed recognition of Mandarin and dialects. To recognize both Mandarin and a specific dialect, create separate projects and configure the corresponding model for each.

   • If no base model meets your needs, set hotwords or create a custom model. For details, see Language Model Customization and Hotword.

     To configure language models and hotwords for a project, go to All Projects in the left navigation pane and click Project Function Configuration to the right of the target project. On the project function configuration page, scroll to the Speech Recognition area and find the Custom Model section. Select existing name hotwords, location hotwords, business-specific hotwords, and language models from the dropdown lists, then click Apply for the changes to take effect.

     To create new hotwords or custom language models, click Create Hotword at the bottom of a hotword dropdown list, or navigate to Hotwords or Custom Linguistic Models under the self-learning platform in the left navigation pane.

Text-to-speech

These steps apply when Project Type is set to text-to-speech only or speech recognition + text-to-speech + speech analysis.

In the text-to-speech (TTS) module, select a text-to-speech model and configure basic parameters such as speech rate, pitch, and volume. Console configurations are bound to the project's appkey when published. If your application does not specify these parameters in the API call, the console defaults apply.

1. Click Project Function Configuration to the right of the target project.

2. In the text-to-speech (TTS) section, click Modify Configuration.

3. Select a voice, then configure speech rate, pitch, and volume in the Basic Parameters section.

   On the text-to-speech only configuration page, select a voice (for example, Zhiyuan) from a voice category on the left (for example, General). In Basic Parameters, set the speech rate, pitch, volume, sample rate, and format.

4. In the Test module on the right, preview the audio.

5. Click Confirm to Use. The configuration takes effect immediately.

Edit a project

Click Edit to the right of the target project to update its name or scenario description.

Delete a project

Click Delete to the right of the target project to remove it.

FAQ

How do I obtain an Appkey, AccessKey ID, and AccessKey Secret?

Appkey is a project-level credential, while AccessKey ID and AccessKey Secret are account-level credentials. They are managed independently.

• Appkey: Sign in to the Intelligent Speech Interaction console. In the left navigation pane, click All Projects. Find the target project in the list and view the Appkey in the Appkey column. You can also click Project Function Configuration to open the project details page, where the Appkey is displayed in the Basic Information section with a copy button.

• AccessKey ID and AccessKey Secret: Sign in to the RAM console and obtain the AccessKey pair on the user security settings page.Sign in to the RAM console and obtain the AccessKey pair on the user security settings page.

Why is my recording file recognition task in a queue, and how do I view transcription results?

Recording file recognition is an asynchronous service. Tasks are queued and processed in order. Under normal conditions, results are returned within 3 hours. Task queuing is expected behavior.

To view the transcription result, sign in to the Intelligent Speech Interaction console, open the offline file transcription task list, find the target task, and click the playback button to view the result.

How do I preview voices in a long-text speech synthesis voice package?

Batch previewing all voices in a purchased voice package is currently not supported. Only selected voices offer online preview.

To preview an available voice, sign in to the Intelligent Speech Interaction console, create or open a project, and click Project Function Configuration. In the speech synthesis service section, click Modify Configuration. Select a voice from the categories on the left, then use the test area on the right to enter text and preview the synthesized audio.