process command is the core of NoteWise. Point it at a YouTube video URL and it fetches the transcript, sends it to your configured LLM, and writes structured Markdown study notes to your output directory.
Basic usage
./output/ by default (or the directory set in your config). Each video gets its own subdirectory named after the video title.
Output structure
After processing, your output directory will contain:Chapter-aware output
For videos longer than one hour that have chapters defined, NoteWise automatically splits the notes into per-chapter files instead of a single document:All process options
--model / -m — select LLM model
Override the default model for this run. Accepts any LiteLLM-format model string.
gemini/gemini-2.5-flash. Set a persistent default via DEFAULT_MODEL in ~/.notewise/config.env.
--output / -o — output directory
Write notes to a specific directory instead of the configured default.
--language / -l — transcript language
Specify the preferred transcript language. Use standard BCP-47 language codes. The flag is repeatable — NoteWise tries each language in order and uses the first available track.
["en"]. Override persistently with DEFAULT_LANGUAGES in ~/.notewise/config.env.
--temperature / -t — LLM sampling temperature
Control how deterministic the LLM output is. Accepts a float from 0.0 (fully deterministic) to 1.0 (most varied). The default is 0.7.
--max-tokens / -k — maximum tokens per response
Cap the number of tokens the LLM can generate per response. Useful when working with models that have lower limits or when you want to control cost. Defaults to the model’s own maximum.
--force / -F — re-process cached videos
By default, NoteWise skips videos it has already processed (detected via the local SQLite cache). Pass --force to rerun the full pipeline regardless.
Use
--force after changing your prompt configuration or switching models to regenerate notes without deleting the cache.--no-ui — plain output for CI and scripts
Disables the Rich live dashboard and prints plain progress lines to stdout instead. Recommended for CI pipelines, cron jobs, and any context where you are piping or redirecting output.
--quiz — generate a quiz alongside notes
In addition to study_notes.md, generate a quiz.md file containing multiple-choice questions based on the video content.
--export-transcript — export the raw transcript
Save the raw transcript to a file. Accepts either txt (plain text) or json (timestamped segments).