Token Estimator & Text Chunker

It uses a characters-per-token ratio, not a model-specific tokenizer. The default of four characters per token is close for English prose; code and non-English text need a lower ratio. Treat the figure as a budget, not an exact bill.

Around 5–10% of the target size works for most prose. Overlap is capped at half the target so the splitter cannot loop. Use zero overlap when you want clean, non-repeating sections.

It tries paragraph boundaries first, then sentence boundaries, and only cuts mid-sentence when a single sentence is larger than the target. This keeps each chunk readable.

No. Token counting and chunking run in the browser. The text, chunk sizes and downloaded files stay on your device.

Yes. Each chunk is plain text with a copy button. Download all chunks with separator headings if you prefer to work through a long document offline.