News

Availability of Chinese models

2019/11/04

We are happy to announce the availability of Mandarin Chinese support for Alpha Cephei. Models both for web API and mobile API are available on Github. Try them here .

Presentation in Novosibirsk State University

2019/04/02

We also gave a talk on VOSK in Novosibirsk State University. The auditory was very warm and welcoming and helped us a lot with precise and interesting questions.

Watch the video

CodeFest 2019 in Novosibirsk, Russia

2019/03/30

AlphaCephei gave a talk on modern speech recognition in CodeFest conference in Novosibirsk, Russia. We were excited to announce VOSK system there and very thankful for the warm reception of the talk. The video will be available soon.

VOSK Speech Recognition System Released

2019/03/20

AlphaCephei is happy to announce the next generation lifelong learning speech recognition system called VOSK. This system is based on audio indexing technology and enables new unique capabilities for speech recognition as well as human-like behavior. Find more details on Github.

Products

Smart Home

Manage your home, lights, locks and other devices with simple voice commands

Many Languages

Chinese, Hindi, Spanish, English, Russian, French, German, all major languages of the world

Chatbots

Build intelligent conversational interfaces with your clients using our ASR and NLP technology

Hearables

Build voice-enabled mobile applications working offline without internet connection, providing security, privacy and smooth fast response experience.

Robots

Make natural language orders to our new mates everywhere - at the office, at factory, on space station

Automotive

Don't distract yourself on the road - manage music and entertainment with voice

Open Source

VOSK is an innovative, next generation open source speech recognition toolkit. It focuses on life-long semi-supervised learning with huge datasets. The idea is to apply audio fingerprinting approach to learn speech sounds as well as other non-speech sounds. We are also considering support for distributed multiparty learning as core feature of the toolkit.

To name few advantages or our approach: we can quickly train on 100000 hours of speech data, we can correct recognizer behavior just by adding samples. We can also parallelize training across thousands of nodes.

You are welcome to pull our code from Github and try it.

About

Alpha Cephei Inc. primary mission is scientific one. We spent more than 10 years researching speech production, speech recognition and all related areas. We have made a significant progress in understanding the nature of such a marvelous communication method as speech.

At the same time we are also the experts in implementation of the state of the art algorithms in practical systems. We excelled our capabilities during numerous projects in many companies around the world.

We hope you'll find our software, our publications and thoughts useful.

Speech Recognition