CLI tool for transcription
- Transcribe audio and output to stdout in json format (supports multiple file input)
- Helps with drafting captions for your editing workflow
When you run prattl prepare
prattl installs a python distribution specific to your OS to your system, This means you don't need to manage Python dependencies or risk corrupting your existing environments.
prattl clean
will completely remove this distribution.
Under the hood, prattl is using distil-whisper, which runs locally on your system. The better your GPU, the faster the transcription, if you do not have a GPU, it will use your CPU. To create a transcription, use the command:
prattl transcribe <filepath/to/audio.mp3>
You can provide multiple file paths, and prattl will transcribe all of them as a single batch. This means the efficiency increases with more files!
Upon completion, the output will be a JSON object. For instance, if you run:
prattl transcribe test1.mp3 test2.mp3 test3.mp3
the output will be:
{
"test1.mp3": "test1.mp3's transcription",
"test2.mp3": "test2.mp3's transcription",
"test3.mp3": "test3.mp3's transcription"
}
- windows/386
- windows/amd64
- darwin/arm64
- darwin/amd64
- linux-gnu/arm64
- linux-gnu/amd64
Note
CUDA architecture GPUs can take advantage of GPU acceleration for transcription
ffmpeg installed and included in $PATH
- Download (or build) binary to chosen location
- Include the path of the binary in
$PATH
- Run
prattl prepare
to install necessary python dependencies to$HOME/.prattl
directory - Prattl is ready to be used!
go installed and included in $PATH
ffmpeg installed and included in $PATH
- Clone repository
- cd into the cloned repository
- Either run
go build
and then manually move the resulting binary into your$PATH
, or just rungo install
to install the executable file to$GOPATH/bin
- Run
prattl prepare
to install necessary python dependencies to $HOME/.prattl directory - You're good to go!
Help with any prattl command
prattl help
Prepare the python distribution at $HOME/.prattl
required by prattl
prattl prepare
Remove the python distribution built at $HOME/.prattl
by prattl
prattl clean
Transcribe provided audio by pathname(s), output transcription to stdout/json
prattl transcribe <filepath/to/audio.mp3> <filepath/to/second_audio.mp3>