Output

Green marks additions; red strikethrough marks deletions. CRLF and LF line endings are normalized; space differences are preserved.

Processed in this browser. No uploads or automatic saving.

How do I compare two prompt versions?

Paste the original and revised prompts and select line or character comparison. Line mode helps review added or removed rules; character mode highlights wording and punctuation edits. Added, removed, and unchanged character counts accompany the diff. Swap the versions to reverse the comparison. One empty input means a complete addition or deletion.

Is a prompt diff a model-quality evaluation?

Text differences help check whether edits omitted output constraints, changed roles, or added context. They do not prove that a revised prompt performs better. This tool does not run a model or invent a quality score. Evaluate accuracy, completeness, and costs separately with the same test cases.

How are line endings, spaces, and Unicode compared?

CRLF and LF line endings are normalized to LF so platform differences do not obscure actual edits. Spaces remain significant. Character mode compares Unicode code points; combining sequences and complex emoji can contain multiple code points. Character counts are not model token counts and cannot directly determine API costs.

Are my prompts uploaded or saved?

No. Both versions are compared in a browser worker without uploads or automatic saving. Each version is limited to 100,000 characters. Very complex differences stop with a request to reduce the input. Download a text record containing both original inputs for manual review; no version history is created automatically.