Updated README

This commit is contained in:
Yuta Hayashibe 2022-09-25 00:46:37 +09:00
parent 5598a06233
commit 6ac27504df

View file

@ -8,10 +8,9 @@
Streaming transcriber with [whisper](https://github.com/openai/whisper).
Transcribing in real time, enough machine power is needed.
## Example
## Setup
```bash
# Setup
git clone https://github.com/shirayu/whisper_streaming.git
cd whisper_streaming
poetry install --only main
@ -19,7 +18,11 @@ poetry install --only main
# If you use GPU, install proper torch and torchaudio with "poetry run pip install -U"
# Example : torch for CUDA 11.6
poetry run pip install -U torch torchaudio --extra-index-url https://download.pytorch.org/whl/cu116
```
## Example of microphone
```bash
# Run in English
poetry run whisper_streaming --language en --model base
```
@ -29,6 +32,15 @@ poetry run whisper_streaming --language en --model base
- ``-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
- ``--debug`` outputs logs for debug
### Parse interval
If you want quick response, set small ``-n`` and add ``--allow-padding``.
However, this may be at the sacrifice of accuracy.
```bash
poetry run whisper_streaming --language en --model base -n 20 --allow-padding
```
## Example of web socket
⚠ **No security mechanism. Please make secure with your responsibility.**