CanDoYa
All tools
Saved toolsYour favorites

Add to favorites

Saved in this browser only.
Private browser utility / Social & Creator

Free Text Similarity Checker

English only

Runs entirely in your browser - no upload, no sign-up.

Live workspaceLocal processing

Paste two texts to compare their semantic similarity.

Your texts are processed in a browser worker and are never uploaded. The first run downloads the public AI model.

Compare two texts by meaning

This tool uses sentence embeddings to measure how similar two texts are in meaning, not just word overlap. Paraphrases and synonyms score high.

The MiniLM model is designed for English text. Results for other languages are only a rough guide.

Share this tool
text similarity / browser utility
01 / Overview

What is a text similarity checker?

A text similarity checker measures how alike two passages are in meaning, not just word overlap. This tool converts each text into a semantic embedding using a local MiniLM model, then compares them with cosine similarity. You get a score from 0-100% and a label (Nearly identical, Highly similar, etc.) without sending your text to a server.

02

How to use

  1. 01
    Paste your texts

    Enter or paste two passages in the Text A and Text B fields.

  2. 02
    Click Compare

    The tool downloads the model on first run, then computes embeddings for both texts.

  3. 03
    Review the score

    A percentage and label show how similar the texts are in meaning. The cosine value is also shown for reference.

  4. 04
    Copy or adjust

    Copy the result or paste different texts to compare again.

03

Who it's for

  • Content teams can check whether two versions of an article or product description convey the same core message.
  • Researchers and students can compare abstracts, summaries, or paraphrases to ensure they capture the original meaning.
  • Translators can verify that a translation preserves the semantic intent of the source text.
  • QA and testing can spot when generated or templated text drifts from the intended meaning.

This tool compares meaning, not words. Unlike a word-diff tool, it understands that 'the cat sat on the mat' and 'a feline was seated on the rug' are highly similar despite sharing few exact words. It uses the Apache-2.0 all-MiniLM-L6-v2 model, a 22 MB sentence encoder trained on English text.

The model runs entirely in your browser. WebGPU uses roughly 30 MB of quantized weights when supported; the WebAssembly CPU fallback uses roughly 54 MB. Browser and model runtime files download separately and may be cached.

FAQ

Is my text uploaded to a server?

No. Both texts are passed to a worker inside your browser, and model inference runs on your device. The worker downloads the public Transformers.js runtime and MiniLM model files from external hosts, but those requests do not contain the text you pasted. CanDoYa does not receive or store it.

Is the text similarity checker free?

Yes. You can compare texts without an account or usage fee. Your browser supplies the computing power, so the first run can take longer while it downloads the AI runtime and model. Cached files often make later runs faster.

How does it measure similarity?

The tool converts each text into a 384-dimensional embedding using the MiniLM model with mean pooling and L2 normalization. It then computes the cosine similarity between the two embeddings, which ranges from -1 (opposite) to 1 (identical). The result is mapped to a 0-100% scale and labeled (Nearly identical, Highly similar, etc.) based on calibrated thresholds.

What does the cosine similarity number mean?

Cosine similarity measures the angle between two vectors. A value of 1.0 means the texts are identical in meaning; 0.0 means they are orthogonal (unrelated); negative values are rare for normalized embeddings. The percentage shown is the cosine value mapped to 0-100%, clamping negative values to 0%.

Why do similar-sounding texts sometimes score low?

The model ranks texts by semantic meaning, not phonetics or surface similarity. Two texts with the same topic but opposite sentiment (e.g., 'I love this' vs. 'I hate this') may score lower than expected because the embeddings capture the full context, not just keywords.

Does it work for languages other than English?

The MiniLM model was trained as an English sentence encoder. Other Latin-script languages may return a score, but the thresholds and accuracy are not calibrated for them. For best results, use English text.

How long can the texts be?

The model can handle texts of any length, but very long passages (>1000 words) may take longer to embed. For best results, compare passages of similar length.

Will it work offline?

It can work after your browser has cached the Transformers.js runtime, tokenizer, and model files, but cache availability is controlled by the browser. A first visit needs a network connection, and private browsing, storage cleanup, or cache eviction can require another download.