40 lines
1,005 B
Markdown
40 lines
1,005 B
Markdown
# pandora_transcribe
|
|
|
|
use whisper_timestamped to add automatic transcriptions to pan.do/ra
|
|
|
|
## installation
|
|
|
|
cd /srv/pandora/pandora
|
|
git clone https://code.0x2620.org/0x2620/pandora_transcribe transcribe
|
|
|
|
add "transcribe" to LOCAL_APPS in local_setttings.py
|
|
|
|
## install whisper_timestamped
|
|
|
|
python3 -m venv venv /opt/whisper-timestamped/
|
|
/opt/whisper-timestamped/bin/pip install \
|
|
torch==1.13.1+cpu \
|
|
torchaudio==0.13.1+cpu \
|
|
-f https://download.pytorch.org/whl/torch_stable.html
|
|
/opt/whisper-timestamped/bin/pip install whisper-timestamped
|
|
|
|
## configuration
|
|
|
|
add a user called subtitles and create 2 lists for that user: Queue and Transcribed
|
|
alternatively see options for `./manage.py transcribe` to use another user or list name
|
|
|
|
## run
|
|
|
|
in a terminal run
|
|
|
|
./manage.py transcribe
|
|
|
|
|
|
## install a service
|
|
|
|
copy systemd/service/pandora-transcribe.service to /etc/systemd/system/pandora-transcribe.service and
|
|
|
|
systemctl enable --now andora-transcribe.service
|
|
|
|
|
|
|