Skip to main content
The CLI uses the same PremoveITN.normalize() implementation as the Python API.

​
Normalize one transcript

Select a device with --device:
Valid values are auto, cpu, mps, and cuda.

​
Process stdin

When the text argument is omitted, the CLI reads newline-delimited transcripts from stdin:
Stdin mode loads the model once and processes each line in order. Use this mode instead of starting a new process for every transcript. Run premove-itn --help for command help or premove-itn --version for the installed package version. Normal output uses stdout. Diagnostics use stderr. The command requires a text argument or piped stdin. A model-load or normalization error stops processing and returns a non-zero exit status. For startup and device limits, see Deployment.