Strip ANSI escape codes from a string

Remove every ANSI / VT escape sequence from text — for log scrubbing, plain-text emails, terminals that don't support color, or pipe-friendly CLI output. The regex below matches CSI, OSC, and standalone ESC sequences across the most common forms.

The regex (Perl-compatible)

PCRE

\x1b(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~]|\][^\x07\x1b]*(?:\x07|\x1b\\))

Matches `ESC` (0x1b) followed by either: a single non-printable two-byte sequence (`ESC X` where X is < 0x40, e.g. RIS `\033c`); a CSI sequence (`ESC [` + parameter bytes + final byte 0x40–0x7e); or an OSC / DCS / SOS / PM / APC sequence (`ESC [ ] P X ^ _ ]` + any bytes + terminator `BEL` or `ESC \`).

By language

Python (stdlib re)

import re

ANSI = re.compile(r'\x1b(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~]|\][^\x07\x1b]*(?:\x07|\x1b\\))')

def strip_ansi(s: str) -> str:
    return ANSI.sub('', s)

JavaScript / Node (no dependencies)

const ANSI = /\x1b(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~]|\][^\x07\x1b]*(?:\x07|\x1b\\))/g;

export function stripAnsi(s) {
  return s.replace(ANSI, '');
}

// npm alternative:
//   npm i strip-ansi
//   import stripAnsi from 'strip-ansi';
//   stripAnsi('\u001b[31mhi\u001b[0m');  // 'hi'

Go (regexp)

package ansi

import "regexp"

var ansiRe = regexp.MustCompile("\x1b(?:[@-Z\\\\-_]|\\[[0-?]*[ -/]*[@-~]|\\][^\x07\x1b]*(?:\x07|\x1b\\\\))")

func Strip(s string) string {
    return ansiRe.ReplaceAllString(s, "")
}

Rust (regex crate)

// Cargo.toml: regex = "1"
use regex::Regex;
use std::sync::OnceLock;

fn ansi() -> &'static Regex {
    static R: OnceLock<Regex> = OnceLock::new();
    R.get_or_init(|| Regex::new(
        r"\x1b(?:[@-Z\\-_]|\[[0-?]*[ -/]*[@-~]|\][^\x07\x1b]*(?:\x07|\x1b\\))"
    ).unwrap())
}

pub fn strip_ansi(s: &str) -> String {
    ansi().replace_all(s, "").into_owned()
}

Bash / shell (sed)

# GNU sed with -E / -r:
sed -E 's/\x1b(\[[0-?]*[ -\/]*[@-~]|\][^\x07\x1b]*(\x07|\x1b\\)|[@-Z\\\\-_])//g' input.log > clean.log

# perl one-liner (more portable across BSD/macOS sed):
perl -pe 's/\e(?:[@-Z\\-_]|\[[0-?]*[ -\/]*[@-~]|\][^\a\e]*(?:\a|\e\\))//g' input.log

C (single-pass byte-level state machine, no regex)

#include <stdio.h>

// Strip every ESC sequence from stdin to stdout.
int main(void) {
    int c;
    while ((c = getchar()) != EOF) {
        if (c != 0x1b) { putchar(c); continue; }
        int next = getchar();
        if (next == EOF) break;
        if (next == '[') {                   // CSI
            int b;
            while ((b = getchar()) != EOF && (b < 0x40 || b > 0x7e)) {}
            continue;
        }
        if (next == ']' || next == 'P' || next == 'X' ||
            next == '^' || next == '_') {    // OSC, DCS, SOS, PM, APC
            int b;
            while ((b = getchar()) != EOF) {
                if (b == 0x07) break;        // BEL terminator
                if (b == 0x1b) { getchar(); break; } // ST = ESC \\
            }
            continue;
        }
        // ESC X (two-byte) — drop both bytes
    }
    return 0;
}

What this does NOT strip

Lone control characters that aren't escape sequences (BEL `\x07`, BS `\x08`, raw `\r` / `\n`, etc.) are preserved by the recipes below — those usually belong in the text. Add a second pass like `s/[\x00-\x08\x0b\x0c\x0e-\x1f]//g` if you want to strip every C0 control too.

Canonical pages for the escape families the regex above matches and removes.

The regex (Perl-compatible)

By language

What this does NOT strip

Related sequences