Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/paulrobello/par_yt2text

Extract metadata, transcripts, and comments from YouTube videos with optional audio transcription using OpenAI Whisper API
https://github.com/paulrobello/par_yt2text

Last synced: about 1 month ago
JSON representation

Extract metadata, transcripts, and comments from YouTube videos with optional audio transcription using OpenAI Whisper API

Host: GitHub
URL: https://github.com/paulrobello/par_yt2text
Owner: paulrobello
License: mit
Created: 2024-10-21T05:33:17.000Z (3 months ago)
Default Branch: main
Last Pushed: 2024-10-29T21:12:19.000Z (2 months ago)
Last Synced: 2024-11-18T16:17:24.527Z (about 2 months ago)
Language: Python
Size: 134 KB
Stars: 2
Watchers: 3
Forks: 0
Open Issues: 0
Metadata Files:
- Readme: README.md
- License: LICENSE

Awesome Lists containing this project

README

# PAR YT2Text

[![PyPI](https://img.shields.io/pypi/v/par_yt2text)](https://pypi.org/project/par_yt2text/)
[![PyPI - Python Version](https://img.shields.io/pypi/pyversions/par_yt2text.svg)](https://pypi.org/project/par_yt2text/)
![Runs on Linux | MacOS | Windows](https://img.shields.io/badge/runs%20on-Linux%20%7C%20MacOS%20%7C%20Windows-blue)
![Arch x86-63 | ARM | AppleSilicon](https://img.shields.io/badge/arch-x86--64%20%7C%20ARM%20%7C%20AppleSilicon-blue)
![PyPI - License](https://img.shields.io/pypi/l/par_yt2text)

PAR YT2Text Based on yt By Daniel Miessler with the addition of OpenAI Whisper for videos that don't have transcripts.

[!["Buy Me A Coffee"](https://www.buymeacoffee.com/assets/img/custom_images/orange_img.png)](https://buymeacoffee.com/probello3)

## Features

- Extract metadata, transcripts, and comments from YouTube videos
- If the transcript is not available, optionally use OpenAI Whisper API to transcribe the audio

## Prerequisites

* To install PAR YT2Text, make sure you have Python 3.11.
* Create a GOOGLE API key
* If you want to use OpenAI Whisper API, create an OPENAI API key (An OpenAI key is not needed for local OpenAI Whisper).

### [uv](https://pypi.org/project/uv/) is recommended

#### Linux and Mac
```bash
curl -LsSf https://astral.sh/uv/install.sh | sh
```

#### Windows
```bash
powershell -ExecutionPolicy ByPass -c "irm https://astral.sh/uv/install.ps1 | iex"
```

## Installation

### Installation From Source WITHOUT support for local OpenAI Whisper

* Clone the repository:
```bash
git clone https://github.com/paulrobello/par_yt2text.git
cd par_yt2text
uv sync
```

### Installation From Source WITH support for local OpenAI Whisper

* Clone the repository:
```bash
git clone https://github.com/paulrobello/par_yt2text.git
cd par_yt2text
uv sync -U --extra local-whisper
```

### Installation From PyPI WITHOUT support for local OpenAI Whisper

```bash
uv tool install par_yt2text
```

```bash
pipx install par_yt2text
```

### Installation From PyPI WITH support for local OpenAI Whisper

To install PAR YT2Text from PyPI with local OpenAI Whisper, run any of the following commands:

```bash
uv tool install -U 'git+https://github.com/paulrobello/par_yt2text[local-whisper]' --index https://download.pytorch.org/whl/cu121 --index-strategy unsafe-best-match
```

```bash
pipx install 'par_yt2text[local-whisper] @ git+https://github.com/paulrobello/par_yt2text' --pip-args="--extra-index-url https://download.pytorch.org/whl/cu121"
```

## Usage
Create a file called `~/.par_yt2text.env` with your Google API key and OpenAI API key in it.
```bash
GOOGLE_API_KEY= # needed for youtube-transcript-api
OPENAI_API_KEY= # needed for OpenAI API whisper audio transcription (An OpenAI key is not needed for local OpenAI Whisper).
PAR_YT2TEXT_SAVE_DIR= # where to save the transcripts if you dont specify a folder in the --save option
```

Whisper audio transcription will only be used if you specify the `--whisper` or `--local-whisper` option and the video does not have a transcript.
If you want to force the use of whisper audio transcription, use the `--force-whisper` option with one of the `--whisper` or `--local-whisper` options.

Often the transcript will come back a single long line.
PAR YT2Text will attempt to add newlines to the transcript to make it easier to read unless you specify the `--no-fix-newlines` option.

### Local Whisper
While the OpenAI Whisper API is fast and inexpensive a free local option is available.
**NOTE: Local whisper mode can be very slow on cpu. If you have a CUDA enabled GPU it will be used unless you specify the `--whisper-device` option.**
`turbo` is the default local model however you should consult the [OpenAI Whisper documentation](https://github.com/openai/whisper?tab=readme-ov-file#available-models-and-languages) to see what models are available and select the best one for your VRAM needs.

### Running from source
```bash
uv run par_yt2text --transcript --whisper 'https://www.youtube.com/watch?v=COSpqsDjiiw'
```

### Running if installed from PyPI
```bash
par_yt2text --transcript --whisper 'https://www.youtube.com/watch?v=COSpqsDjiiw'
```

### Example of forcing use of local Whisper if tool was installed with local Whisper enabled
```bash
par_yt2text --transcript --force-whisper --whisper-local 'https://www.youtube.com/watch?v=COSpqsDjiiw'
```

### Options
```
usage: par_yt2text [-h] [--duration] [--transcript] [--comments] [--metadata] [--no-fix-newlines] [--whisper] [--local-whisper]
[--whisper-device {auto,cpu,cuda}] [--force-whisper] [--whisper-model WHISPER_MODEL] [--lang LANG] [--save FILE]
url

positional arguments:
url YouTube video URL

options:
-h, --help show this help message and exit
--duration Output only the duration
--transcript Output only the transcript
--comments Output the comments on the video
--metadata Output the video metadata
--no-fix-newlines Dont attempt to fix missing newlines from sentences
--whisper Use OpenAI Whisper to transcribe the audio if transcript is not available
--local-whisper Use Local OpenAI Whisper to transcribe the audio if transcript is not available
--whisper-device {auto,cpu,cuda}
Device to use for local Whisper cpu, cuda (default: auto)
--force-whisper Force use of selected Whisper to transcribe the audio even if transcript is available
--whisper-model WHISPER_MODEL
Whisper model to use for audio transcription (default-api: whisper-1, default-local: turbo)
--lang LANG Language for the transcript (default: English)
--save FILE Save the output to a file
```

## Whats New
- Version 0.2.0:
- Added support for local OpenAI Whisper
- Version 0.1.0:
- Initial release

## Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

## License

This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

## Author

Paul Robello - [email protected] (Based on yt By Daniel Miessler)