Speech to text

[SpaceShaman]

This AI function allows you to record your voice and use it as a prompt. You can use this function in combination with text to speech

stt-2024-10-04_18.27.15.mp4

Combination with text to speech

stt-tts-2024-10-04_18.40.30.mp4

How to use

1. Install the requirements

   pip install marvin[audio]

2. Add OpenAI api key to system environment (for marvin's use)

   export OPENAI_API_KEY='<your-key>

3. Save below code as speech_to_text.py in the ~/.config/shell_gpt/functions directory.

import subprocess
import sys

from instructor import OpenAISchema
from pydantic import Field


class Function(OpenAISchema):
    """
    Record phrase from microphone and return it as text.
    """

    text: str = Field(
        ...,
        descriptions="Recorded text from microphone.",  # type: ignore
    )

    class Config:
        title = "speech_to_text"

    @classmethod
    def execute(cls) -> str:
        process = subprocess.Popen(
            [
                sys.executable,
                "-c",
                "import marvin.audio; print(marvin.transcribe(marvin.audio.record_phrase()))",
            ],
            stderr=subprocess.DEVNULL,
            stdout=subprocess.PIPE,
        )
        stdout, _ = process.communicate()
        return stdout.decode()