How to use
- Paste or open text; spelling, grammar and style are checked on your device as you type.
- Click a highlight or a suggestion to accept, ignore it, or add a word to your dictionary (kept in this browser).
- Optionally press Download model and Check with grammar model for a sentence-by-sentence model pass, then Apply all fixes or accept them one by one.
Worked example
“Their going to the libary tomorow.” flags libary → library and tomorow → tomorrow (spelling). With the optional model, “He dont know were it is.” gets two suggestions: dont → doesn't and were → where.
Supported formats and limits
| Input | Text up to 20,000 characters, TXT, Markdown, HTML or DOCX files |
|---|---|
| Output | Corrected text, Suggestions with inline highlights |
| Limits | Spelling and rule checks run on your device with no download beyond the site's own 0.5 MB dictionary. The optional grammar model (Flan-T5-base) is a one-time download of about 279 MB from Hugging Face, cached afterwards. Your text never leaves this tab. |
| Engine | nspell with the en_US Hunspell dictionary (dictionary-en) in a Web Worker; conservative rule checks (repeated words, a/an, capitals, spacing, your/you're, its/it's, then/than, could of, passive voice, long sentences); optional Flan-T5-base (Apache-2.0) via transformers.js, one sentence at a time |
Limitations
- Rule checks only catch clear patterns; many grammar errors (like “Their going”) need the optional model, and it misses some too.
- Spelling uses an English (US) dictionary; names and technical terms may be flagged until you add them.
- Model suggestions can be wrong or change your meaning; read each one.
Questions
What does it catch without the model?
Spelling against an English (US) dictionary, and clear patterns such as repeated words, a versus an, missing capitals, spacing around punctuation and common confusions (your/you're, its/it's, then/than, could of). Passive voice and long sentences get style hints without automatic fixes. Many grammar errors need the model.
How reliable is the grammar model?
Flan-T5-base is a small open model (Apache-2.0, about 279 MB, downloaded once from Hugging Face). Its suggestions are machine-generated: it misses many errors and sometimes rewrites correct sentences or changes the meaning, so read each one before accepting it.
Is my text uploaded?
No. Checks run in this tab. The only network request is the optional model download, which fetches model files and does not include your text. Words you add to your dictionary are kept in this browser.
Privacy
On-device model. Processing runs in this browser. The open-source model files are downloaded once from the model host (Hugging Face) and cached; your content is not uploaded.
- Hugging Face: Only after you press Download model: requests for the model files (Flan-T5-base, about 279 MB, only if you use the optional grammar model) from huggingface.co and its CDN, which see your IP address. Your text and files are never sent.
See the privacy policy for how toolsdocks handles data.