second pass · beta

Rescrub the leftovers.

Paste or upload a file that already went through de-identification but still contains real values. Every remaining name, number, room, ID, and note is replaced with a deterministic fake. Table structure, DDL, and referential joins stay intact.

options

Salt (drives deterministic fakes — same salt = same fakes)

Preserve PK-shaped IDs (id, *_id, *_key, uuid)Preserve NULL / empty valuesPreserve booleans and enum-like tokens

Drop a .sql or .csv file, or click to pick

Streams multi-GB files through a local Web Worker — nothing is uploaded.

Or paste text (≤ 5 MB)

dump.sql

SQL · no output yet

output preview

Drop a file or paste text above, then click Rescrub.

how it works

Deterministic fakes. Every value is hashed with your salt, so users.id = 42 becomes the same fake wherever it appears — foreign-key joins keep working across tables.
Schema-safe. Only literal values inside INSERT tuples, Postgres COPY blocks, and CSV cells are rewritten. Column names, DDL, comments, and row counts pass through byte-for-byte.
Local-only. Files stream through a Web Worker in this tab. Nothing is uploaded.
Not realistic. Fakes are drawn from a small deterministic pool — a name becomes a plausible-looking name, but not one tied semantically to the original. This is a test-data scrubber, not a synthetic-data generator.

← First-pass de-identify Knowledge baseWorks offline — download the standalone build for zero network.