ViewCSV
Open the original CSV and check how its bytes become text before changing the separator. This tool reports a Unicode byte-order mark, tests UTF-8 validity and lets you compare other input encodings. Review the parsed columns, confirm the result, then download UTF-8 CSV. Files stay in your browser; the tool never guesses missing characters.
Repair an import in three checks
- Open the original CSV, TSV or TXT. Choose the symptom, or try the synthetic Windows-1252 sample to see the complete workflow. Pasted text is already decoded and cannot reveal its original byte encoding.
- Select the input encoding and delimiter, then choose Preview selected settings. For the sample, select Windows-1252; check that Café is readable, 00731 retains its zeros and the quoted note remains one cell.
- Choose the output delimiter and whether to include the UTF-8 BOM. Review formula protection, check the confirmation box, then download the repaired CSV. Open in editor transfers the decoded table for further editing.
What the encoding and separator checks establish
A UTF-8 or UTF-16 BOM identifies the declared Unicode encoding. Without a BOM, valid UTF-8 bytes do not uniquely identify the source encoding. Candidate snippets show several possible readings of the beginning; they are not confidence scores. Selecting an encoding validates the complete file strictly, without silently inserting replacement characters.
The separator comparison examines up to 20 complete records from the first 65,536 text characters. It respects quoted separators and quoted line breaks. A column count is only a clue: confirm the actual headers and cells. For layout questions without an encoding problem, use Change CSV delimiter and encoding.
Save a reviewed UTF-8 copy
Downloads use UTF-8, optional BOM, CRLF record endings and necessary CSV quoting. Choose comma, semicolon, tab or pipe. Existing numbers, dates, spaces and identifiers stay text. Formula protection is enabled initially and prefixes formula-like values with an apostrophe; turn it off only when you want those original values in the output.
The preview shows normalization of blank or duplicate headers and padding of shorter records. Warnings disclose these changes. An unfinished quoted field stops parsing; this tool does not invent its missing boundary. To keep long identifiers as text in a workbook instead, continue with CSV to Excel; for a delimited text download, use CSV to TXT.
When the original characters cannot be recovered
Characters already replaced by � or question marks cannot be reconstructed from the damaged copy. Use the original export when available. Pasting loses the original encoding evidence, and choosing UTF-8 output does not repair a wrong input decoding. UTF-32 is not supported. This interactive workflow supports up to 25 MiB, 100,000 data rows, 250 columns and 2 million cells; decoding also respects the existing text-size limit. No data is uploaded or sent to analytics.
Example: readable accents and a note kept in one cell
Synthetic Windows-1252 file, after choosing the correct input encoding. The output uses commas and UTF-8 with BOM; its invisible BOM is included below. The note's embedded CRLF stays inside quotes.
sku;name;note 00731;Café;"red, blue next; line"
sku,name,note 00731,Café,"red, blue next; line"
Questions and answers
No. A BOM is evidence for the declared Unicode encoding. Without one, several decodings may be possible. Compare real names and characters, select the input encoding and confirm the table before exporting.
The receiving application may expect a different separator. Compare the input columns here, then choose its required output separator. You can also use the application's text import controls to specify the delimiter explicitly.