Package transcribe.py as installable CLI #3

New issue

Closed

opened 2026-09-01 20:37:56 +00:00 by Mark · 0 comments

Mark commented

2026-09-01 20:37:56 +00:00

Owner

Currently a single ad-hoc script at /tmp/whisperx-smoke/transcribe.py. Needs to live in the repo with:

pyproject.toml declaring deps (whisperx, yt-dlp, ffmpeg-python)
Entry point: transcribe command on PATH after uv tool install or pipx install
Sensible CLI surface: transcribe URL_OR_FILE [--model M] [--lang CODE] [--out PATH] [--align] [--diarize] [--batch-size N] [--device cuda|cpu] [--compute-type float16|int8]
Error handling: clean messages for each known failure mode (network, model not cached, OOM, format unsupported)

Scope: 1-2 files, additive, no external infra.

Currently a single ad-hoc script at `/tmp/whisperx-smoke/transcribe.py`. Needs to live in the repo with: - `pyproject.toml` declaring deps (whisperx, yt-dlp, ffmpeg-python) - Entry point: `transcribe` command on PATH after `uv tool install` or `pipx install` - Sensible CLI surface: `transcribe URL_OR_FILE [--model M] [--lang CODE] [--out PATH] [--align] [--diarize] [--batch-size N] [--device cuda|cpu] [--compute-type float16|int8]` - Error handling: clean messages for each known failure mode (network, model not cached, OOM, format unsupported) **Scope:** 1-2 files, additive, no external infra.