Output formats

Every list command renders through the same formatter, so what you learn for search applies to video, channel, comments, and the rest. Pick a format with -o, --output, narrow the columns with --fields, or reshape each row with --template.

Choosing a format

By default the format is auto: a human-readable table when ytb is writing to a terminal, and JSONL when the output is piped or redirected. That way reading in a shell and piping into another program both do the right thing with no flag.

Override it with -o:

ytb search example -o table    # a rounded, color-aware grid for reading
ytb search example -o markdown # a GitHub pipe table to paste into docs (alias: md)
ytb search example -o json     # a single JSON array
ytb search example -o jsonl    # one JSON object per line, for piping
ytb search example -o csv      # comma-separated, spreadsheet friendly
ytb search example -o tsv      # tab-separated
ytb search example -o url      # just the canonical URL, one per line
ytb search example -o id       # just the id, one per line
ytb search example -o raw      # the raw underlying value

table, markdown, csv, and tsv carry a header row; pass --no-header to drop it. json emits one array for the whole result, while jsonl emits one object per line so it streams and composes with tools like jq -c. markdown escapes any pipe in a cell, so a title like Go in 100 Seconds | Fireship stays a valid table.

Color

On an interactive terminal the table and json/jsonl formats are colored: the table draws a dim rounded border with an accented bold header, and JSON keys, strings, numbers, and literals are highlighted. Color is suppressed the moment output is not a terminal, so a pipe always gets plain, parseable bytes and ytb ... | jq is never affected. Force the choice with --color always|never (or set NO_COLOR). markdown, csv, tsv, url, id, and raw are never colored, so they stay safe to redirect into a file. A too-wide table shrinks to fit the terminal instead of wrapping at the edge.

Selecting columns with --fields

--fields takes a comma-separated list of table column names. These are the short, lowercase names you see in the table header, not the JSON keys:

ytb search "rust tutorial" -n 100 --fields title,channel,views
ytb video dQw4w9WgXcQ --fields title,channel,views -o table

The column names differ from the snake_case keys in JSON output. For example the table column is views, while the same value in -o json is view_count. Use the column names with --fields and the JSON keys when you parse JSON downstream.

Templating with --template

--template is a Go text/template applied once per row. Templated fields use the Go struct field names, which are capitalized, so views on the table becomes {{.Views}} here:

ytb search example --template '{{.Title}} ({{.Views}})'
ytb video dQw4w9WgXcQ --template '{{.ID}} {{.Title}} {{.Channel}}'

A template wins over --fields and -o, since you are describing the exact line to print.

Piping ids and urls

-o id and -o url exist for chaining. Any command that takes an id or URL also accepts - to read a list from stdin, so the output of one command feeds the next:

ytb search "go programming" -o id | ytb video -
ytb playlist PLFgquLnL59alCl_2TQvOiD5Vgm1hCaGSI -o id | ytb video -

The first command emits one id per line, and ytb video - resolves each to full metadata.