| layout | title | permalink |
|---|---|---|
default |
FAQ |
/faq |
Kaldi is a research speech recognition toolkit which implements many state of the art algorithms. Vosk is a practical speech recognition library which comes with a set of accurate models, scripts, practices and provides ready to use speech recognition for different platforms like mobile applications or Raspberry Pi. If you are doing research, Kaldi is probably your way. If you want to build practical applications with plug and play library, consider Vosk.
Vosk reuses best practices for accurate speech recognition from many other toolkits, not just Kaldi. In our research we use Nvidia Nemo, Fairseq and many other open source libraries, our goal is to build life-long learning platform which continuously improves speech recognition for major languages and use cases.
Stay tuned!
We train our models on thousands hours of speech data, they should be pretty good out of box. Still, if you look for better accuracy contact us, we will try to help you.
Try to reproduce the problem with a file recording and share it with us, we will check.
The process of building a new language model consists of the following steps:
- Data collection (you can collect audiobooks with text transcription
from projects like
librivox, transcribed podcasts, or setup web data collection. - Data cleanup
- Model training
- Testing
We can help you with the steps since we are interested to support as many languages as possible. Feel free to contact us.