Skip to content

Latest commit

 

History

History
72 lines (48 loc) · 3.2 KB

README.md

File metadata and controls

72 lines (48 loc) · 3.2 KB
Raw

Pytranscript 🎙️

Pytranscript is a powerful Python library and command-line tool designed to seamlessly convert video or audio files into text and translate them into various languages. It acts as a simple yet effective wrapper around Vosk, ffmpeg, and deep-translator, making the transcription and translation process straightforward.

Prerequisites

Before using pytranscript, ensure you have the following dependencies installed:

  • ffmpeg for audio conversion.
  • vosk-models required for speech recognition. You will have to specify to your specific model path in the --model argument.

Installation

pip install pytranscript

Usage

Command Line

pytranscript INPUT_FILE [OPTIONS]

Options

  • -m, --model - Path to the Vosk model directory. Always required.
  • -o, --output - Output file where the text will be saved. Default: input file name with .txt extension.
  • -f, --format - Format of the transcript. Must be one of 'csv', 'json', 'srt', 'txt', 'vtt' or 'all'. Default: input file extension.
  • -li, --lang_input - Language of the input / the model. Default: auto.
  • -lo --lang_input - Language to translate the text to. Default: no translation.
  • -s, --start - Start time of the audio to transcribe in seconds.
  • -e, --end - End time of the audio to transcribe in seconds.
  • --max_size - Will stop the transcription if the output file reaches the specified size in bytes. Takes precedence over the --end option.
  • --keep-wav - Keep the converted audio wav file after the process is done.
  • -v, -verbosity - Verbosity level. 0: no output, 1: only errors, 2: errors, info and progressbar, 3: debug. Default: 2.

Example

The most basic usage is:

pytranscript video.mp4 -m vosk-model-en-us-aspire-0.2 -lo fr -f srt

Where vosk-model-en-us-aspire-0.2 is the Vosk model directory. The text will be translated from English to French, and the output will be saved in video.srt.

Using the keep-wav option can be useful if you want to do many transcriptions within the same file, allowing you to use the same .wav file for each transcription, thus saving conversion time. ⚠️ The .wav file is cropped according to the start and end time options.

API

The API provides a Transcript object containing the time and text. The translate method can be used to get another Transcript object with the translated text. The output saved in a file in the cli is just a method to_{format} of the Transcript object.

A reproduction of the previous example using the API:

import pytranscript as pt

wav_file = pt.to_valid_wav("video.mp4", "video.wav", start=0, end=None)
transcript = pt.transcribe(wav_file, model="vosk-model-en-us-aspire-0.2", max_size=None)
transcript_fr, errors = transcript.translate("fr")

transcript_fr.write("video.srt")

Contributing

Contributions are welcome! For major changes, please open an issue first to discuss what you would like to change. Tests can be run with pytest. Use ruff with ruff format . to format the code before committing.