Prepare the input
Paste the older text on the left field and the intended newer text in the revised field. Normalize irrelevant whitespace first when spacing differences should not create line changes.
Compare two versions line by line
Text Diff
Build a bounded line-level longest-common-subsequence comparison with unchanged, added, and removed lines, entirely in this browser tab.
- 01Private by designYour text stays in this browser tab.
- 02Bounded and responsiveUp to 250,000 characters and 1,500 lines on each side.
- 03Ready to reuseCopy the result or download a branded TXT file.
Counting and transformations run locally. Nirmion does not send the text in this workspace to an API.
Product guide
Compare versions with bounded line alignment
Nirmion finds a longest common line subsequence, then walks both versions to show retained context, removals, and additions. It compares exact line strings and does not infer moved blocks, semantic similarity, or word-level edits.Choose the rule
There are no fuzzy settings. Exact equal lines become context; unmatched original lines are marked with minus and unmatched revised lines with plus. The bounded algorithm rejects overly large line sets rather than freezing the tab.
Review the output
Read clusters of removal and addition as possible modifications, then compare important wording in the source. Do not apply the output as a Git patch: it has labels and prefixes but no hunks, file paths, or patch guarantees.
How it works
Longest common subsequence over exact lines
Line endings are normalized and a typed matrix records the longest common subsequence length for every bounded pair of positions. Reconstruction emits unchanged, added, or removed lines while preserving source order. Each side is capped at 1,500 lines and 250,000 characters to bound memory and CPU use; output is capped at one million characters.
Useful when
Inspect small and medium text revisions
Line comparison provides a quick private preflight before a deeper editorial or version-control review.
- Compare release notes, policies, prompts, configuration snippets, transcripts, or documentation drafts without uploading them.
- Confirm that a revised checklist or generated text retained expected lines and introduced only intended additions.
- Create a readable comparison artifact for discussion before moving source files into a formal review workflow.
Questions answered
Text Diff questions
Practical boundaries matter with text tools. These answers explain what is measured, what is retained, and when a manual review is still needed.
Is this a word-by-word diff?
No. Equality and metrics operate on complete lines. A one-character edit makes the old line a removal and the new line an addition.
Why is there a 1,500-line limit?
Longest-common-subsequence alignment has quadratic worst-case work. The explicit bound protects browser responsiveness and memory. Use version-control or dedicated diff software for larger files.
Can I apply the output as a patch?
No. It resembles a simple diff for readability but does not include standardized hunks, paths, hashes, or application guarantees.
Privacy and retention
Your draft stays on your device.
Nirmion performs these operations with browser JavaScript. The workspace does not call the catalogue API with your source text, and completed-job content is kept only in this component memory for re-download.
A downloaded file remains in your browser downloads, device storage, backups, and any location where you share it. Clear or close the page when working on a shared device, and avoid pasting secrets into extensions or third-party clipboard managers.
Tool details
A private bounded line diff for draft comparison
Nirmion Text Diff compares original and revised plain text using exact line-level longest-common-subsequence alignment. It reports unchanged, added, and removed lines plus processing time, helping users inspect drafts, notes, policies, prompts, and configuration snippets without sending them to an API.
Each side is limited to 250,000 characters and 1,500 lines, and the readable output downloads as text_nirmion_tools_diff.txt. The page clearly separates this artifact from semantic analysis, word diffing, and standards-compliant version-control patches.