Efficient speech recognition software with Russian language support?

Question

Speech recognition

Can you recommend something that works? There is, for example, a recording of a radio program, I want to make a more or less suitable transcript.

Answer 1

At the moment there is no solution.
The Russian language is phonetically too complex. It is possible to achieve more or less correct recognition of individual words, but one can only dream of a normal transformation of a dictated sentence into text.

Answer 2

The recording of a radio program can be recognized using Voco Professional www.speechpro.ru/product/rechevye-produkty/voco/sp...
Download the demo first and make sure that the quality of your recordings is sufficient for recognition.

Answer 3

This question was already asked by me just a month ago. =)
True, they didn’t offer anything particularly sensible =(

Answer 4

google speech API

Answer 5

Speech Pro. (was higher) They have a saas solution recently, it's better to write to them. It didn't work for us, but it might work for you.
Google speech Api
Nuance has a couple of decent products: www.nuance.com/for-individuals/by-product/dragon-for-pc/index.htm

Answer 6

Immediately the main question: you for the money or indulge. There are a lot of cloud solutions with acceptable recognition, but they are paid.

Answer 7

The update to Evernote for Android has a nice Russian speech-to-text dictation.

Answer 8

hardly credit to Evernote
Evernote for Android Update: Speech-to-text and Big Widget Enhancements

The speech-to-text feature uses Google Android's text transcription service. You must be online for it to work. Also, this capability is available on version 4 of the Android Operating System (aka Ice Cream Sandwich) and some devices with version 2.2 and above. If you see the speech-to-text option appear when you update, then you'll know that your phone or tablet supports the functionality.

Answer 9

Try the Wax open source speech recognition library:
https://github.com/alphacep/vosk-api
Benefits of the library:

• Supports 9 languages ​​- Russian, English, German, French, Portuguese, Spanish, Chinese, Turkish, Vietnamese. Others will be added soon
  
• Works without network access even on mobile devices - Raspberry Pi, Android, iOS
  
• Installed with a simple pip3 install vosk command without additional steps
  
• Models for each language are only 50MB, but there are also much more accurate large models for more accurate recognition.
  
• Made for streaming sound processing, which allows you to implement instant response to commands
  
• Supports several popular programming languages ​​- Java, C#, Javascript
  
• Allows you to quickly adjust the recognition dictionary to improve recognition accuracy
  
• Allows you to identify the speaker
  

To recognize radio broadcasts, you can install a server from docker, the project is here
https://github.com/alphacep/vosk-server
Runs with one command:

docker run -d -p 2700:2700 alphacep/kaldi-en:latest