What is Intelligent Speech Interaction

Updated at:

Intelligent Speech Interaction (ISI) uses technologies such as speech recognition, speech synthesis, and natural language understanding to add natural voice interaction capabilities to applications. It supports scenarios such as intelligent Q&A, automated quality inspection, real-time court transcription, live speech captions, and interview transcription, and is used in industries such as finance, justice, and e-commerce.

Note

Intelligent Speech Interaction 2.0 is now available. The self-learning platform and other tools improve speech recognition accuracy. The new version also provides an enhanced console and a more user-friendly SDK. Activate the service to get started.

Short sentence recognition

Short sentence recognition transcribes audio clips of up to 1 minute. It is suitable for short interactions such as voice search, voice commands, and voice messages, and can be integrated into apps, smart home appliances, and intelligent assistants. For more information, see Short sentence recognition API reference.

Real-time speech recognition

Real-time speech recognition transcribes audio streams of any length and returns text as speech is received. It includes intelligent sentence segmentation and returns the start and end times of each sentence. It is suitable for live video captions, real-time meeting transcription, real-time court transcription, and intelligent voice assistants. For more information, see Real-time speech recognition API reference.

Recording file recognition

Recording file recognition transcribes uploaded audio files. It is suitable for call center quality inspection, court trial database entries, meeting summaries, and medical record entry. For more information, see Recording file recognition API reference.

Note

Free trial users receive transcription results within 24 hours. Paid users receive results within 3 hours, except when more than 500 hours of recordings are uploaded within 30 minutes. For large-scale transcription requirements, contact an Alibaba Cloud pre-sales representative.

Speech synthesis

Speech synthesis uses deep learning to convert text into natural, fluent speech. Multiple voices are available, and speech rate, pitch, and volume can be adjusted. It is suitable for intelligent customer service, voice interaction, audiobook narration, and accessibility announcements. For more information, see Speech synthesis API reference.

CosyVoice speech synthesis

CosyVoice speech synthesis is based on large-scale pretrained language models and deeply integrates text understanding with speech generation. It accurately interprets diverse text and converts it into natural speech that closely resembles a human voice.

Offline speech synthesis

Offline speech synthesis uses an on-device model to convert text into natural, fluent speech in poor or unavailable network conditions.

Multiple voices are available, and speech rate, pitch, and volume can be adjusted. The service is suitable for in-car navigation, smart hardware, audiobook narration, and accessibility announcements. It is integrated by using an SDK and supports multiple hardware platforms. Billing is based on the number of activated devices, providing flexible cost control. For more information, see Offline speech synthesis API reference.

Self-learning platform

The self-learning platform improves speech recognition accuracy by training hotwords and custom language models. Add hotwords and use hotword models to improve recognition results. In fields such as justice and finance, custom language models can improve recognition accuracy for domain-specific speech.

What's next

  • Getting started: Get started with Intelligent Speech Interaction.

  • Pricing: Learn how Intelligent Speech Interaction is billed.

  • Developer guide: Learn key terms and how to obtain an access token.

  • Console user guide: Learn about the features available in the console.

  • API reference: Choose a service, such as short sentence recognition, real-time speech recognition, recording file recognition, or speech synthesis.

  • Self-learning platform: Improve recognition accuracy by training hotwords and custom language models.

  • Best practices: Learn how to implement Intelligent Speech Interaction in common scenarios.

  • FAQ: Find solutions to common issues.