How to dictate a grep pattern, and why you probably should not

Shell regex is the worst dictation target in this set. A grep invocation combines flags, quotes and a pattern, and VoiceGem returns grep - E"^ error"log.txt for a simple case — three separate whitespace problems in a single short command.

A nine-word grep invocation runs about 10 seconds typed against about 3.3 seconds spoken. (sources: Dhakal, Feit, Kristensson & Oulasvirta, “Observations on Typing from 136 Million Keystrokes”, CHI 2018, 168,000 participants, 136 million keystrokes; Yuan, Liberman & Cieri, “Towards an Integrated Understanding of Speaking Rate in Conversation”, Interspeech 2006, 2,438 American English telephone conversations (Switchboard corpus). The speech figure is conversational rate, not a measured dictation rate — no such measurement exists.)

Say this

What lands in your editor

grep - E"^ error:"application.log

This tool runs entirely in your browser. Nothing you type is sent anywhere, stored, or logged.

Step by step

1
Say the command name
Speak “grep” or “rg” directly. Command names are not table phrases, so they pass through untouched.
2
Expect the flag to detach
Say “dash E” and accept the space. The hyphen attaches on neither side because its VoiceGem table entry serves subtraction, so every shell flag arrives detached.
3
Say the quoted pattern
Speak “double quote”, the pattern, “double quote”. The quotes bind on both sides, so they glue to the flag before and the filename after.
4
Type it instead
Write the command by hand. This page's honest recommendation is that shell regex is not worth dictating, and a page that hid that would be selling rather than describing.

Shell regex combines every weakness the Deterministic Pass has

Three separate problems land in one short command. Flags use hyphens, which detach; paths use slashes, which detach; and quotes bind on both sides, which glues them to the flag before and the filename after. A nine-word grep invocation needs edits in three different places.

Naming the limit is the useful thing to do here. VoiceGem is built for code that is mostly identifiers with punctuation between them, and a shell command is the opposite — mostly punctuation with short fragments in between. There is no phrasing that fixes it and no per-language mode that would, because a mode would mean detecting your context.

Where doing it by hand breaks down

Nowhere, for the command itself. Shell commands are what keyboards are for, every developer has the flags in muscle memory, and this page is not going to invent a case.

Long patterns inside the quotes are the only part worth speaking. A pattern with several classes and quantifiers is the same dense punctuation it is in any language, and dictating just that fragment into the VoiceGem tool on this site, then pasting it between the quotes, is the workflow that actually helps.

Doing this somewhere else? speech-to-code converter, in one place, how dictating Bash functions works, how dictating Go regular expressions works, or dictating Java regular expressions specifically.

Common questions

Generally no. Flags, paths and quotes each produce a whitespace problem, so a short command needs edits in three places.

The pattern inside the quotes, on its own. Dictate that fragment into the browser tool, strip the spaces, and paste it into the typed command.

No. A per-language mode would require detecting your context, which is the guess the Deterministic Pass exists to avoid.

Dictate just the pattern fragment and paste it into your typed command.

Open the tool