Supprime 38 doublons. Coupe 1247 cellules. Standardise 89 dates. Vous dit exactement ce qui a changé.
Nettoyeur en ligne gratuit pour données Excel et CSV — supprimer les doublons, couper les espaces, normaliser dates et nombres, corriger les caractères invisibles, supprimer les lignes vides. Chaque nettoyage est livré avec un rapport de transparence par action et un classeur de différences optionnel. Sans inscription. Jusqu'à 50 Mo par fichier.
Bâti sur la même classe de technologie de traitement de documents d'entreprise utilisée dans les flux critiques d'institutions financières Fortune 100 — gratuit, sans compromis sur la qualité.
- 🔒 256-bit HTTPS
- ⏱ 1-hour Auto-Delete
- ✅ GDPR Compliant
- 👁 No Human Review
- 🤖 No AI Training on Your Data
- 📊 Per-action Transparency Report
Déposez votre fichier Excel ou CSV ici, ou cliquez pour parcourir
Max 50 Mo · xlsx, xlsm, xls, xlsb, ods, csv, tsv
Pourquoi « nettoyé » signifie souvent « boîte noire »
La plupart des nettoyeurs gratuits Excel vous donnent un bouton de téléchargement et c'est tout. Vous ne voyez pas ce qui a été supprimé, ce qui a été réécrit, ni si l'outil a silencieusement fusionné deux clients qu'il n'aurait pas dû. Le nettoyeur d'ExcelMaster est bâti sur du traitement de documents de niveau entreprise — chaque nettoyage vient avec un rapport de transparence par action et un classeur de différences optionnel. Vous voyez, avant de valider, exactement combien de doublons nous supprimerons, combien de cellules nous couperons, et quelles dates nous réécrirons.
Ce qui est nettoyé
Exact dedupe + optional AI fuzzy dedupe
Drop full-row duplicates, or scope to specific columns ("same customer regardless of timestamp"). Optionally let Claude catch near-duplicates that exact match misses — "John Smith" vs "Smith, John", "Acme Inc" vs "Acme Incorporated".
Whitespace and invisible chars
Trim leading/trailing spaces, collapse runs of internal whitespace, strip NBSP / zero-width space / BOM / control chars that came in via copy-paste from web pages or PDFs.
Date standardisation
Recognises MM/DD/YYYY, DD-MM-YYYY, ISO, "May 7 2026", "2026年5月7日", "2026년 5월 7일" and rewrites them all to one target format. ISO by default — sorts correctly as text, machine-readable everywhere.
Number standardisation
Strips currency symbols ($ € ¥ £ ₹ R$ kr…) and thousands separators (en "1,234.56" / de "1.234,56" / fr "1 234,56"). Output is a real Excel number you can SUM, AVERAGE, chart.
Case transformation
UPPER, lower, Title Case or Sentence case across the whole sheet or scoped to specific columns ("Title Case the Customer Name column only").
Blank row & column removal
Removes rows where every cell is empty after trimming. Optionally remove blank columns too. Header row is always preserved.
Comment nettoyer des données Excel en ligne
- 1Upload your Excel or CSV file — drag & drop, max 50 MB, no signup required.
- 2Click Preview impact — we estimate how many rows, cells and dates we'd touch, before you commit. Industry-first transparency.
- 3Pick cleaning actions — dedupe, trim, normalise dates, normalise numbers, change case, remove blanks. All toggles default to conservative settings.
- 4Click Clean now — typically under 3 seconds for 10k rows. Download the cleaned file plus the optional side-by-side diff workbook so you can verify every change.
Pourquoi ExcelMaster surpasse les alternatives
Per-action transparency report
Industry standard is the black-box "your file is cleaned" success card. We tell you exactly: 38 duplicate rows removed (3 by fuzzy AI), 1247 cells trimmed, 89 dates standardised, 23 invisible chars stripped. Audit-grade.
Side-by-side diff workbook
Optional second .xlsx: original vs cleaned rows, changed cells highlighted yellow, plus a "Cleaning Summary" sheet. Verify every cleaning decision before accepting. No competitor ships this in a free tool.
AI fuzzy dedupe — opt-in
Catches "John Smith" vs "Smith, John", "Acme Inc" vs "Acme Incorporated". Powered by Claude Sonnet 4.6 via our LLM Proxy. Conservative — only merges when confident. Capped at 500 rows for cost.
11 languages, 145 date locales
Recognises Chinese (2026年5月7日), Japanese (令和8年5月7日 / 2026年5月7日), Korean (2026년 5월 7일), German (07.05.2026), French (07/05/2026), and more. Same engine that powers our PDF→Excel multilingual extraction.
Header row protected
We never trim, normalise, or case-change your header row. Your column titles arrive verbatim — the convention every spreadsheet tool gets wrong eventually.
Free — no compromises on output quality
Same engine the $49/month commercial tools use. We've made it free with no row caps, no watermarks, no signup. Funded by our paid Excel AI Add-in for desktop — most users never need to upgrade.
Questions fréquentes
How is this different from Excel's built-in "Remove Duplicates"?
Excel's built-in dedupe only does exact-match comparison and offers no transparency. We add fuzzy AI dedupe (catches "John Smith" vs "Smith, John"), per-action transparency report, side-by-side diff workbook, and 6 other cleaning actions in one pass. And it's free — Excel's commercial-grade equivalent (Power Query) requires Office 365.
Will fuzzy AI dedupe wrongly merge two distinct entities?
We tuned the prompt to be conservative — "better to miss a near-duplicate than to wrongly merge two distinct entities". The AI is asked to skip merges it isn't confident about. The opt-in toggle is off by default. Even when on, the side-by-side diff workbook lets you verify every merge before accepting the cleaned file. If you spot a wrong merge, you can rerun without the fuzzy toggle.
What date formats are recognised?
MM/DD/YYYY, DD/MM/YYYY (with - / . separators), ISO 8601, "May 7 2026" / "7 May 2026", Chinese (2026年5月7日), Japanese, Korean (2026년 5월 7일), German (7. Mai 2026), French (7 mai 2026), and Excel's native serial number format. We use the .NET DateTime parser with 8 culture fallbacks. Anything that ambiguously parses is left untouched.
What invisible characters do you strip?
NBSP (U+00A0), zero-width space (U+200B), zero-width non-joiner (U+200C), zero-width joiner (U+200D), BOM (U+FEFF), word joiner (U+2060), and all control characters U+0000–U+001F except tab (U+0009), newline (U+000A), and carriage return (U+000D). These are the chars that creep in via copy-paste from PDFs and web pages.
Is there a row limit?
Up to Excel's native row limit per sheet (1,048,575 data rows). We use a streaming load when source files exceed 200,000 cells to avoid memory pressure. The fuzzy AI dedupe pass is capped at 500 rows for cost reasons — we'll surface a warning when your sheet exceeds that and skip just the AI pass while still doing exact dedupe.
Will my data leave the server?
Files are processed on our servers in EU and uploaded over TLS. Cleaning happens entirely in our infrastructure. The optional fuzzy AI dedupe pass sends row signatures (the cells you asked us to compare) to Claude via our LLM Proxy — that's the only outbound data leg. Files auto-delete from our servers within 1 hour. We do not train on your data.
Does it work for CSV?
Yes. CSV and TSV are first-class inputs — most "clean my CSV online" SEO traffic arrives via messy-csv use cases. Output can be CSV, xlsx, xlsm, xls, xlsb, or ods. CSV output flattens to the first sheet only.
Can I clean a password-protected workbook?
Yes. Provide the open password and we clean as usual. We never brute-force, we never share, we never log the password. (For workbooks where you've lost the password, see our Excel Unlock tool.)
What's the difference between "Remove blank rows" and "Remove duplicates"?
Blank-row removal drops rows where every cell is empty after trimming. Dedupe drops rows where every cell value matches another row. They are independent toggles — you can run either alone or both together.
Can I scope dedupe to specific columns?
Yes. By default we compare the full row, but you can pick specific column indices — "two rows are duplicates iff these specific columns match". Useful for "same customer (col 0,1,2) regardless of timestamp in col 5". The same scoping applies to the optional fuzzy AI pass.
Why is the cleaned file slightly smaller / larger than the source?
Smaller is usually because we removed blank rows / duplicates / strip currency symbols. Slightly larger can happen when you turn on AI fuzzy dedupe + diff workbook generation — the diff is a separate file. The transparency report's "source rows" vs "output rows" tells you exactly how many rows were dropped.
Do you keep my Excel formatting / formulas?
Cell values are cleaned; cell styles and number formats are preserved. Formulas are preserved when the cell type stays the same (text → text). When we normalise a string-typed currency cell into a real number, the formula on dependent cells stays intact and now sees a real number — that's usually what you want. The diff workbook lets you verify case-by-case.