g3n-data
Support繁中
← All tools

Text Cleaner

Combine six editable presets and Unicode, duplicate-word, line, markup and formatting rules while retaining the original.

text-cleanerLOCAL WORKSPACE

1. Local input and settings

Inputs stay in memory: no uploads, URLs, storage or requests. Each field32KiB; full output1MiB. Worker:10s startup/1s processing. Editing cancels work and clears old exports. Use synthetic examples for learning.

Six editable presets
Composable cleanup rules
Remove duplicate words

Only unprotected text is deduplicated, before line cleanup. Intl.Segmenter (und, word) finds Unicode words; segmentation can vary by runtime. First spelling, punctuation and spacing are retained; other enabled rules can still alter them. Case-insensitive comparison uses Unicode lowercase, not full case folding or accent normalization. Code, inline code, Markdown links/images/references and HTML tokens are excluded from both matching and removal. Explicit HTML/Markdown stripping can expose text for cleanup; code remains protected. Results may leave extra spaces or empty lines.

Model, bounds and sources

Each input32KiB/5,000 lines; Markdown/HTML5 nodes5,000/depth30. Code, inline code, HTML and reference/link destinations are protected from character cleanup; explicit TAB/trailing-space options also affect code. Explicit Markdown stripping/unwrap removes syntax and retains code contents. TAB width1–16 (default4); NBSP normalization is independent. Order: special spaces/TAB, per-line trim, space collapse, blank lines, then line-break keep/space/remove. Protected code keeps line breaks; fullwidth only U+FF01–FF5E/U+3000. NFC/NFD mutually exclusive. Emoji uses grapheme properties; invisible removal may change language/emoji. Formatting heuristics do not detect AI authorship. Diff previews first8,192 UTF16 units/200 lines only; full output1MiB. Mixed EOL with structural editing needs explicit LF/CRLF.

Source controls and references inspected2026-10-19.

2. Complete result

Review settings, then process locally.

How to use this tool

Enter local input, choose the explicit rules or algorithm, then run and export the complete result.

When not to use it

Each input32KiB/5,000 lines; Markdown/HTML5 nodes5,000/depth30. Code, inline code, HTML and reference/link destinations are protected from character cleanup; explicit TAB/trailing-space options also affect code. Explicit Markdown stripping/unwrap removes syntax and retains code contents. TAB width1–16 (default4); NBSP normalization is independent. Order: special spaces/TAB, per-line trim, space collapse, blank lines, then line-break keep/space/remove. Protected code keeps line breaks; fullwidth only U+FF01–FF5E/U+3000. NFC/NFD mutually exclusive. Emoji uses grapheme properties; invisible removal may change language/emoji. Formatting heuristics do not detect AI authorship. Diff previews first8,192 UTF16 units/200 lines only; full output1MiB. Mixed EOL with structural editing needs explicit LF/CRLF.