Compare commits

..

2 Commits

Author SHA1 Message Date
gauthiier
16ea900e1b verbose 2024-04-24 10:11:10 +02:00
gauthiier
9f5fe78bc0 small model 2024-04-24 10:10:53 +02:00
2 changed files with 28 additions and 2 deletions

View File

@ -1,5 +1,31 @@
# 🗣️ ✍️ 💾 # 🗣️ ✍️ 💾
```
STT terminal utility based on openai-whisper and pyaudio STT terminal utility based on openai-whisper and pyaudio
```
git clone --recurse-submodules <git-repo>
```
```
usage: ./transcriptum.sh [action]
where action can be: [install, clean, run]
```
```
usage: transcribe.py [-h] [--model {tiny,base,small,medium,large}] [--rms RMS]
[--record_timeout RECORD_TIMEOUT]
[--phrase_timeout PHRASE_TIMEOUT] [--dynamic_threshold]
TRANSCRIPTUM
options:
-h, --help show this help message and exit
--model {tiny,base,small,medium,large}
Whisper model
--rms RMS RMS (energy) threshold for microphone to detect
--record_timeout RECORD_TIMEOUT
Timeout for the microphone recording
--phrase_timeout PHRASE_TIMEOUT
Silence timeout between phrases
--dynamic_threshold Use dynamic rms threshold?
``` ```

View File

@ -21,7 +21,7 @@ state = State.IDLE
def main(): def main():
p = argparse.ArgumentParser(description="TRANSCRIPTUM") p = argparse.ArgumentParser(description="TRANSCRIPTUM")
p.add_argument("--model", default="medium", help="Whisper model", choices=["tiny", "base", "small", "medium", "large"]) p.add_argument("--model", default="small", help="Whisper model", choices=["tiny", "base", "small", "medium", "large"])
p.add_argument("--rms", default=1000, help="RMS (energy) threshold for microphone to detect", type=int) p.add_argument("--rms", default=1000, help="RMS (energy) threshold for microphone to detect", type=int)
p.add_argument("--record_timeout", default=8, help="Timeout for the microphone recording", type=float) p.add_argument("--record_timeout", default=8, help="Timeout for the microphone recording", type=float)
p.add_argument("--phrase_timeout", default=2, help="Silence timeout between phrases", type=float) p.add_argument("--phrase_timeout", default=2, help="Silence timeout between phrases", type=float)