Skip to workspace
PixelOkay
Data Cleaning StudioFrom messy spreadsheets to usable data
Local workspace
Tools
Your data stays on this device.Files are processed locally. No account, upload or tracking is needed by this tool. Download exports to keep a durable copy.
PIXELOKAY / LOCAL DATA TOOLS

Data Cleaning Studio

Inspect issues, build a reusable cleaning recipe, and export the exact result you preview.

Original rows—
Cleaned rows—
Columns—
Empty cells—
03 / DATA PREVIEW

Cleaned table Not loaded

Your clean dataset starts here

Import a file or open the sample to see the rows and columns.

No rows to preview
How it works1 Import a table2 Add cleaning steps3 Inspect the result4 Export locally
Ready to import CSV, TSV, JSON or NDJSON data.

Clean CSV files online without uploading data

PixelOkay Data Cleaning Studio can standardize column names and values, find and replace text, remove duplicate records, fill blank values, standardize missing-value markers, split or join columns, normalize whitespace or numbers, remove empty rows, and filter or sort CSV data. Preview a cleaning pipeline before downloading a UTF-8 CSV or tab-separated file. Reopen an editable project or apply a saved cleaning recipe to another CSV with matching columns.

CSV import, export, privacy and limitations

Supports comma, tab, semicolon and pipe delimited UTF-8 files, with UTF-16 files supported when they include a byte-order mark. Quoted fields may contain delimiters, double quotes and multiple lines. Exports can preserve original cell text or apply a spreadsheet-oriented formula mitigation; no option is universally safe for every spreadsheet program and data pipeline. The tool is not a database, an AI data repair system or a substitute for checking changes before using the results.

Export your cleaned data

Export the entire cleaned table, not only the visible preview rows. JSON/NDJSON always writes strings exactly and does not apply CSV formula mitigation. The source file is unchanged.

Prefixes potentially active spreadsheet formulas with a tab within quoted cells; this modifies those values. Uncheck for exact text fidelity in machine-to-machine pipelines, but do not open untrusted raw values in spreadsheets.

Reusable cleaning recipe

Recipes contain only transformation rules and column names — not your rows. They work with another dataset when required columns exist.

Save project separately to preserve source rows and all cleaning steps in one file.

Working with your data

Import: choose CSV/TSV, including multiline quoted fields; JSON arrays of records and NDJSON are supported with strict numeric guards. The CSV import rejects malformed quote structures and extra fields rather than dropping data.

Clean: choose a step and its column. Split replaces the source column with two columns; join adds a new column and preserves the inputs. Missing markers may be standardized to empty cells. Operations are applied top-to-bottom. Steps can be disabled, deleted, or reordered; invalid pipelines are rejected without replacing the previous result.

Undo: Ctrl/Cmd+Z and Ctrl/Cmd+Shift+Z (or Ctrl+Y) when not typing in a text field. Saves download to your device; recipe recovery is browser-local and may disappear if browser storage is cleared.

Exports: use spreadsheet-safer export for human-facing files and fidelity export for machine pipelines. Spreadsheet CSV formula mitigation is best-effort, not a universal guarantee.

Quality report: downloads JSON containing aggregate per-column missing/distinct counts and step impacts, without source cell values. Large unique sets still require browser memory.

Limits: up to 16 MB, 80,000 data rows, 150 columns, a combined 2,000,000 cells, 160,000 characters per cell and 60 steps. No merged files or Excel XLSX input. JSON numeric decimals and unsafe integers must be supplied as quoted strings.

Using Data Cleaning Studio

Import a supported CSV or TSV file, review its delimiter and header options, then build a cleaning recipe. Compare original and processed previews, inspect column summaries and export the table with an editable project backup.

Budgets include 16 MiB input, 80,000 records, 150 columns, two million total cells and 60 recipe steps; these caps are not all simultaneously reachable. Browser recovery keeps the recipe, not source rows. Quality reports include filenames and column names. Spreadsheet-oriented formula handling and exact-text export differ; neither guarantees safety across every spreadsheet application.

Choose cleaned table data, reusable rules or a project

Practice: review a fictional text-cleaning recipe

  1. Import a small fictional CSV/TSV table with inconsistent spacing in a text column.
  2. Add a supported trimming or normalization step and compare original and processed previews.
  3. Inspect the export settings, then keep the original file, project and any reusable recipe for their separate purposes.

This is an authored cleaning exercise, not a transformed dataset or a data-quality certification.

Data cleaning questions

Will browser recovery restore my source rows?

No. Recovery retains the recipe rather than source rows. Keep the original file and an appropriate downloaded project independently; a recipe alone is not a dataset backup.

Are exact-text and spreadsheet-oriented exports identical or universally safe?

No. Formula handling can change how text is written for spreadsheet use. Review the selected mode and destination application; neither option guarantees safety in every spreadsheet program.

Related tools