Cubic Speech Recognition Demo Cubic ASR Demo SDK Docs Blog Contact

Cobalt's Cubic technology powers all our speech-to-text recognition projects. This demo uses a variant of Cubic that offers general purpose speech recognition. We can work with you to customize Cubic for your unique requirements.

clear Microphone input failed

  • This demo requires Firefox 56.0 or later, or Chrome 62.0 or later. Chrome requires this page to be opened via HTTPS.
  • Please make sure you allow this page to access your microphone when your browser prompts you to do so.
  • You may continue to use the file upload feature.

Loading...

Unable to initialize this demo. Please try again later or contact us for assistance!

This demo is restricted in some ways. Unless specified as telephony or far field, all the models expect 16 kHz audio recorded with a close microphone. If you upload a file with a different sample rate, your results will be less accurate. Please get in touch with us if you have concerns about processing speed or accuracy.
The audio you upload may be stored on our servers for internal purposes; do not include any sensitive information.

Results
star_border

State of the art

Cubic uses deep-learning for fast, accurate speech recognition. Our technology builds upon Kaldi, the leading ASR toolkit in the research community.

cloud_queue

Embedded or On-prem

Cubic is written in C++, and is designed to be CPU and memory efficient. You can integrate Cubic into your application with a C API, or deploy Cubic in your secure cloud with our gRPC API.

attach_money

Affordable

We offer the Cubic runtime and models suitable for your requirements at very competitive rates.You can license based on volume or at fixed rates, whichever best fits your business model.


build

Customized

Our speciality is to build custom Cubic models that are most suitable for your needs. Our machine learning tools can use your data to build the most accurate recognition models for your domain.

cast

Streaming Recognition

Cubic is designed to efficiently handle streaming of large audio files. Instead of waiting for the whole file to process, Cubic starts streaming results as soon as they are produced.

translate

Supported Languages

Cubic supports recognition in several languages, including English (US/UK), Japanese, Spanish, French, Portuguese. We also specialize in building custom models in whatever language you need.