Commands

Commands

Transcription

lwt transcribe <file>
lwt t <file>                         # alias

Flag	Description
--language, -l	Source language or auto-detect
--prompt, -p	Context prompt
--model, -m	tiny, base, small, medium, large-v3
--format, -f	txt, srt, vtt, json
--output, -o	Output file
--task	transcribe (default) or translate (English only)
--diarize	Speaker identification — HuggingFace Diarization
--num-speakers	Hint for diarization
--speaker-names	Comma-separated names
--hf-token	HuggingFace token
--translate-to	Translate via Ollama
--clean	Clean transcript via Ollama
--summarize	Meeting summary via Ollama
--ollama-model	Ollama model for post-processing

Clean existing transcript

lwt clean <transcript>
lwt clean meeting.txt -o meeting.clean.txt

Setup

lwt setup
lwt onboard                            # alias
lwt setup --model medium
lwt setup --quick
lwt setup --model small --force

Models

lwt models list
lwt models status
lwt models download small
lwt models download medium --force

Install

lwt install cuda                     # CUDA 12 cuBLAS + cuDNN
lwt install cuda --toolkit           # + NVIDIA CUDA Toolkit (Windows)
lwt install diarization              # pyannote.audio
lwt install check

Ollama

lwt ollama status
lwt ollama list
lwt ollama pull llama3.2

Config

lwt config show
lwt config set whisper.model medium
lwt config set diarization.hf_token hf_xxx
lwt config path
lwt config reset

See Configuration for all keys.