Intelligent Speech Interaction (ISI) is a cloud-based speech AI service that converts audio to text, synthesizes text into natural-sounding speech, and understands spoken language. It is built on speech recognition, speech synthesis, and natural language understanding technologies. Integrate ISI into your application to transcribe real-time audio streams, process uploaded recordings, and generate lifelike voices. ISI has been applied to many fields such as finance, justice, and e-commerce.
ISI V2.0 is now available, with an updated SDK and a console that includes the self-learning platform for improving speech recognition accuracy. You are welcome to activate Intelligent Speech Interaction.
Choose a capability
ISI provides the following capabilities. Choose based on your audio type and use case.
| Capability | Best for | Turnaround |
|---|---|---|
| Short sentence recognition | Streaming audio up to 1 minute | Real-time |
| Real-time speech recognition | Live streaming audio of any length | Real-time |
| Recording file recognition | Pre-recorded audio files (batch) | 3 hours (paid) / 24 hours (free trial) |
| Speech synthesis | Generating audio from text | Real-time |
| Speaker customization (Enterprise) | Custom TTS voices for your brand | Enterprise |
Short sentence recognition
Short sentence recognition transcribes audio clips up to 1 minute. It suits short-interaction scenarios where responses must be immediate.
Use cases: voice search, voice command, voice short message
Integrations: mobile apps, smart home devices, smart voice assistants
For API details, see Short sentence recognition overview.
Real-time speech recognition
Real-time speech recognition transcribes audio streams of any length as they arrive, producing text output word by word. The built-in intelligent sentence breaking feature detects the start and end time of each sentence automatically.
Use cases: live video subtitles, meeting and court trial recording, real-time smart voice assistants
For API details, see Real-time speech recognition operation notes.
Recording file recognition
Recording file recognition processes audio files you upload asynchronously and returns transcribed text. Turnaround time depends on your edition.
| Edition | Turnaround after upload |
|---|---|
| Free trial | Within 24 hours |
| Paid | Within 3 hours |
If you upload a large volume of audio — for example, more than 500 hours of recordings within 30 minutes — processing may take longer than the standard turnaround. For large-scale batch transcription, contact your Alibaba Cloud pre-sales representative.
Use cases: call center quality monitoring, court trial archival, meeting minutes, medical records
For API details, see Recording file recognition API reference.
Speech synthesis
Speech synthesis converts text into natural-sounding, fluent audio using deep learning. Choose from multiple speakers and adjust speed, intonation, and volume to match your application's tone.
Use cases: intelligent customer service, voice interaction, audiobook narration, accessibility broadcasting
For API details, see Speech synthesis overview.
Speaker customization (Enterprise)
Speaker customization lets you train high-performance text-to-speech (TTS) voices at a fast speed using a small amount of audio data. Custom voices can be used in the ISI console or deployed to smart devices.
To get started with speaker customization, email nls_support@service.aliyun.com.
Self-learning platform
The self-learning platform improves speech recognition accuracy for domain-specific vocabulary and language patterns. It provides two features:
Hotword training: Add industry-specific terms as hotwords to boost their recognition accuracy.
Custom linguistic models: Upload business-specific corpus to train a language model tuned to your domain.
The self-learning platform is especially useful for industries with specialized terminology, such as justice and finance.
What's next
| Resource | Description |
|---|---|
| Getting started | Activate ISI and run your first transcription |
| Pricing | Understand billing for each capability |
| Developer guide | Get an access token and integrate ISI via SDK |
| Console user guide | Explore the ISI console features |
| Documentation of a speech service | Learn how to use a specific speech service such as short sentence recognition, real-time speech recognition, recording file recognition, or speech synthesis |
| Self-learning platform | Train hotwords and custom linguistic models |
| Best practices | Production tips for common ISI use cases |
| FAQ | Answers to common questions |