Speech recognition using Playground

1. In the management console, select the folder you are going to use to work with SpeechKit.

2. Go to SpeechKit.

3. In the left-hand panel, select SpeechKit Playground.

4. Navigate to the Speech recognition tab.

5. Under Recognition parameters:

   • Language: Select the language or leave Automatic.
   • Text normalization: Presents dates and times in numerical format, converts numbers from text to digits, and provides access to additional settings.
   • Profanity filter: Masks profanity.
   • Literature text: Adds capital letters and punctuation marks.
   • Speaker recognition: Attributes each recognized phrase to a particular speaker.
   • Grouping speaker phrases: Divides phrases into two groups by speaker.

6. Click Select file or drag the audio file to the loading area.

   Tip

   Convert the file to a supported audio format beforehand: MP3, WAV, or OGG with the OPUS audio codec. Maximum file size: 60 MB.

7. Classifiers: Finds phrases of a given category in the text, e.g., greetings, negative or obscene language. This works only for Russian.

8. Result processing: Processing of results with the help of an LLM:

   • Model: Select a model for processing. The processing cost depends on the model you select.
   • Instructions:
     • Enter a prompt in the input field or select a ready-made one.
     • Result format: Specify your preferred recognition result format.
     • Add instructions: Add another instruction. You can add up to five instructions in total.

9. Click Start recognition to start speech recognition for the audio file.

   Recognition may take from a few seconds to a few minutes depending on the audio file size.

10. Click View code to get the request code for Python REST or Python gRPC.

SpeechKit Playground features basic speech recognition options. For more flexible recognition settings, use the API.