Skip to content

Speech Recognition

ASR model can rapidly and accurately convert voice input into text, providing robust support for applications such as speech recognition, automatic transcription, and intelligent assistants. Whether it is for meeting minutes, customer service, real-time captions, or in fields like education and healthcare, the speech-to-text model demonstrates immense potential. It not only enhances work efficiency but also significantly improves user experience, making information access and processing more convenient.

The following models are now available for purchase:

  • MaaS Whisper
  • MaaS-AFast-asr
  • MaaS-Arealtime-asr
  • MaaS-ASpeech-Translation

MaaS Whisper

  • Multilingual and Multitask Support

    Capable of handling multiple languages, including English, Chinese, and 97 other languages, it can perform various tasks such as speech recognition, translation, language detection, and voice activity detection without the need for separate training or adjustments for each language or task.

  • High Accuracy

    The accuracy of English speech recognition reaches human levels, even surpassing human performance in some noisy environments. The accuracy for multilingual speech recognition and speech translation is also high, delivering outstanding performance even in zero-shot scenarios.

MaaS-AFast-asr

The basic functions of fast transcription include: speech recognition, language identification, and speaker separation. It can be used for transcribing audio files and synchronous playback, with a speed that is faster than real-time audio. It can be used when quick access to the content of audio recordings is needed, for example:

  • Fast audio recording and subtitles: Quickly obtain the entire transcription of a video or audio file at once.

  • Video translation: If there is audio in different languages, new subtitles can be obtained for the video immediately.

Examples of typical application scenarios

Video subtitle transcription

  • Film and video subtitle production: Use fast transcription to quickly obtain a set of subtitles for the entire video, such as media entertainment subtitles and multilingual subtitles.

  • Course assistance: Quickly generate subtitles for online education videos to help hearing-impaired students better understand course content, while also assisting non-native speakers in keeping up with the course.

  • Video conference assistance: For video conferences with participants speaking different languages, real-time subtitles can be generated for the meeting video, helping attendees better understand the spoken content and facilitating cross-language communication and collaboration.

  • Short video subtitle generation: Subtitles can be quickly generated for short videos or vlogs shared by users on social media, enhancing user experience and allowing viewers to browse content in silent playback environments.

Recording transcription

  • Market research: Quickly transcribe recordings of interviews with market research subjects into text for organization and analysis.

  • On-site press conferences: For different on-site press conferences, transcribe the recordings from the event quickly to organize and publish news.

  • Medical record documentation and diagnosis: Automatically convert medical records into electronic documents for easier storage and retrieval.

  • Learning material organization: Quickly transcribe audio materials from educational lectures, academic reports, etc., into text to facilitate students and teachers in organizing study materials, writing notes, and conducting academic research.

  • Meeting minutes and organization: After a meeting, obtain a written record of the meeting audio to improve the efficiency and accuracy of meeting minutes, making it easier for participants to review the meeting content and implement decisions made.

  • Customer service quality inspection: Quickly listen to and transcribe a large volume of customer service call recordings, helping businesses better understand customer needs, identify issues, assess agent performance, and improve service quality.

  • Content review assistance: For a large amount of user-generated audio content, quickly transcribing it into text facilitates content review, ensuring that published content complies with platform regulations and relevant laws and regulations.

MaaS-Arealtime-asr

The basic functions of real-time transcription include real-time speech recognition and language identification. This involves instant transcription of real-time audio input and generating intermediate results.

Examples of typical application scenarios

Real-time audio transcription

  • Clinical assistance: Doctors can quickly and accurately record patient conditions, symptoms, diagnoses, and other information through speech recognition, improving medical efficiency and reducing the time spent on handwritten or typed records, allowing them more time to focus on patient care.

  • Meeting assistant: In various settings such as meetings, seminars, training lectures, and business negotiations, real-time speech recognition can transcribe the speeches of participants into text in real time.

  • Smart home and Internet of Things: Users can control smart home devices, such as smart speakers, smart TVs, smart air conditioners, and smart curtains, through voice commands. Real-time speech recognition technology accurately recognizes user voice commands and executes the corresponding actions promptly.

  • Instant messaging applications: In social chat software and input methods, the real-time speech recognition feature allows users to input messages via voice. The system automatically converts speech into text in real time for sending, facilitating quick communication for users who find it inconvenient to type, thereby improving communication efficiency.

  • Telephone customer service communication: During phone calls with customers, real-time speech recognition can instantly convert the customer's speech into text, enabling customer service representatives to quickly and accurately understand customer issues and needs, thus providing answers and solutions more efficiently.

MaaS-ASpeech-Translation

The basic functions of MaaS-ASpeech-Translation include language identification, real-time conversion of speech to text, and translation.

Advanced features include:

  • Speech-to-Speech Translation: Combining TTS (Text-to-Speech) capabilities, the voice service also offers the option to read aloud the translated text, allowing for a natural output of the input speech.

  • Multilingual Voice Translation: Multilingual voice translation allows for multiple languages to be present in the same conversation, translating them all into the same target language. There is no need to restart the session when changing the input language or performing any other actions.

  • Multiple Target Language Translation: When outputs in multiple languages are required, the voice service can directly provide functionality to translate the input language into two target languages. Note: If more than two output languages are needed, additional fees will apply for using a separate translation service.

Examples of typical application scenarios

Speech-to-Text Translation

  • Event Live Streaming: In scenarios such as international event live streaming, audio is converted into text in real time for live subtitles, enhancing the effectiveness of the broadcast.

  • International Video Conferences: In video and phone conference settings, the speech of each participant is recorded and translated in real time, preventing the omission of important content and accurately and efficiently producing multilingual meeting records.

  • International Customer Service Centers: Customer service call records are converted into text in real time for call quality inspection and customer information synchronization, also providing raw material for data mining.

Speech-to-Speech Translation

  • Tour Guide Services: When tour guides explain attractions to foreign tourists, they can use speech-to-speech translation devices to convert their explanations into the tourists' native language in real time. This helps tourists better understand the historical and cultural background, features, and other information about the attractions, enhancing their travel experience.

  • Multilingual Hotline Services: Multilingual hotlines established by public service organizations or businesses, such as tourism inquiry hotlines or emergency rescue hotlines, can utilize speech-to-speech translation technology to provide timely and effective services to callers speaking different languages, ensuring accurate communication of information and proper resolution of issues.

Multilingual Translation

  • Travel Translation: When traveling abroad, multilingual speech translation can create a solution that allows customers to translate any input audio into the local language or from the local language back into their native language. This enables them to communicate with locals and better understand their surroundings.

  • Business Meetings: In meetings with participants who speak different languages, multilingual speech translation allows attendees to communicate naturally, as if there were no language barriers.

Translation into Multiple Target Languages

  • Simultaneous Interpretation: Real-time interpretation of multiple languages during international conferences or press releases