whispering/README.md

127 lines
4.1 KiB
Markdown
Raw Normal View History

2022-09-23 10:20:11 +00:00
# Whispering
2022-09-23 10:20:11 +00:00
2022-09-25 15:25:44 +00:00
[![MIT License](https://img.shields.io/apm/l/atomic-design-ui.svg?)](LICENSE)
2022-10-09 13:12:32 +00:00
![Python Versions](https://img.shields.io/badge/python-3.8%20%7C%203.9%20%7C%203.10-blue)
2022-09-25 15:25:44 +00:00
[![CI](https://github.com/shirayu/whispering/actions/workflows/ci.yml/badge.svg)](https://github.com/shirayu/whispering/actions/workflows/ci.yml)
[![CodeQL](https://github.com/shirayu/whispering/actions/workflows/codeql-analysis.yml/badge.svg)](https://github.com/shirayu/whispering/actions/workflows/codeql-analysis.yml)
[![Typos](https://github.com/shirayu/whispering/actions/workflows/typos.yml/badge.svg)](https://github.com/shirayu/whispering/actions/workflows/typos.yml)
2022-09-23 10:20:11 +00:00
2022-09-24 00:38:40 +00:00
Streaming transcriber with [whisper](https://github.com/openai/whisper).
2022-09-24 19:49:17 +00:00
Enough machine power is needed to transcribe in real time.
2022-09-23 10:20:11 +00:00
2022-09-24 15:46:37 +00:00
## Setup
2022-09-23 10:20:11 +00:00
```bash
2022-10-21 13:11:08 +00:00
pip install -U git+https://github.com/shirayu/whispering.git@v0.6.3
2022-09-23 13:19:53 +00:00
2022-09-25 15:25:44 +00:00
# If you use GPU, install proper torch and torchaudio
2022-10-03 13:56:11 +00:00
# Check https://pytorch.org/get-started/locally/
2022-09-24 03:38:37 +00:00
# Example : torch for CUDA 11.6
2022-09-25 15:14:20 +00:00
pip install -U torch torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
2022-09-24 15:46:37 +00:00
```
2022-10-09 15:53:09 +00:00
If you get ``OSError: PortAudio library not found`` in Linux, install "PortAudio".
```bash
sudo apt -y install portaudio19-dev
```
2022-09-24 15:46:37 +00:00
## Example of microphone
2022-09-24 02:02:40 +00:00
2022-09-24 15:46:37 +00:00
```bash
2022-09-24 00:26:37 +00:00
# Run in English
whispering --language en --model tiny
2022-09-23 10:20:11 +00:00
```
2022-09-24 00:38:40 +00:00
- ``--help`` shows full options
2022-10-15 04:36:30 +00:00
- ``--model`` sets the [model name](https://github.com/openai/whisper#available-models-and-languages) to use. Larger models will be more accurate, but may not be able to transcribe in real time.
- ``--language`` sets the language to transcribe. The list of languages are shown with ``whispering -h``
- ``--no-progress`` disables the progress message
2022-09-29 18:06:41 +00:00
- ``-t`` sets temperatures to decode. You can set several like ``-t 0.0 -t 0.1 -t 0.5``, but too many temperatures exhaust decoding time
2022-09-23 13:46:27 +00:00
- ``--debug`` outputs logs for debug
- ``--vad`` sets VAD (Voice Activity Detection) threshold. The default is ``0.5``. 0 disables VAD and forces whisper to analyze non-voice activity sound period
2022-10-03 14:27:54 +00:00
- ``--output`` sets output file (Default: Standard output)
2022-09-23 13:19:53 +00:00
2022-09-24 15:46:37 +00:00
### Parse interval
2022-10-15 03:56:08 +00:00
By default, whispering performs VAD for every 3.75 second.
This interval is determined by the value of ``-n`` and its default is ``20``.
When an interval is predicted as "silence", it will not be passed to whisper.
If you want to disable VAD, please make VAD threshold 0 by adding ``--vad 0``.
2022-10-02 14:01:52 +00:00
2022-10-15 03:56:08 +00:00
By default, Whisper does not perform analysis until the total length of the segments determined by VAD to have speech exceeds 30 seconds.
However, if silence segments appear 16 times (the default value of ``--max_nospeech_skip``) after speech is detected, the analysis is performed.
2022-10-15 03:56:08 +00:00
2022-09-24 12:54:25 +00:00
## Example of web socket
2022-09-24 12:57:46 +00:00
**No security mechanism. Please make secure with your responsibility.**
2022-09-24 12:54:25 +00:00
2022-09-24 18:59:09 +00:00
Run with ``--host`` and ``--port``.
2022-09-24 15:48:27 +00:00
2022-09-24 15:55:37 +00:00
### Host
2022-09-24 12:54:25 +00:00
```bash
whispering --language en --model tiny --host 0.0.0.0 --port 8000
2022-09-24 12:54:25 +00:00
```
2022-09-24 15:48:27 +00:00
### Client
2022-09-24 12:54:25 +00:00
```bash
2022-09-29 13:16:51 +00:00
whispering --host ADDRESS_OF_HOST --port 8000 --mode client
2022-09-24 12:54:25 +00:00
```
You can set ``-n`` and other options.
2022-09-24 18:59:09 +00:00
2022-10-17 13:23:50 +00:00
## For Developers
1. Install [Python](https://www.python.org/) and [Node.js](https://nodejs.org/)
2. [Install poetry](https://python-poetry.org/docs/) to use ``poetry`` command
3. Clone and install libraries
```console
# Clone
git clone https://github.com/shirayu/whispering.git
# With poetry
poetry config virtualenvs.in-project true
poetry install --all-extras
poetry run pip install -U torch torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
# With npm
npm install
```
4. Run test and check that no errors occur
```bash
poetry run make -j4
```
5. Make fancy updates
2022-10-21 13:08:43 +00:00
6. Make style
```bash
poetry run make style
```
7. Run test again and check that no errors occur
2022-10-17 13:23:50 +00:00
```bash
poetry run make -j4
```
2022-10-21 13:08:43 +00:00
8. Check typos by using [typos](https://github.com/crate-ci/typos). Just run ``typos`` command in the root directory.
2022-10-17 13:23:50 +00:00
```bash
typos
```
2022-10-21 13:08:43 +00:00
9. Send Pull requests!
2022-10-17 13:23:50 +00:00
2022-09-23 10:20:11 +00:00
## License
- [MIT License](LICENSE)
- Some codes are ported from the original whisper. Its license is also [MIT License](LICENSE.whisper)