Text Chunker guide
Split a passage into measured character or word groups and inspect every boundary before using the pieces.
What this tool does
The Text Chunker divides input by a selected maximum number of characters or by an exact number of whitespace-separated words. Character mode can prefer word boundaries. If a single uninterrupted token is longer than the maximum, that token is split so the stated character limit remains meaningful. Turning word preservation off cuts the character sequence at exact intervals instead.
Word mode normalizes whitespace inside each result and groups the requested number of non-whitespace tokens. Character mode counts Unicode code points rather than UTF-16 storage units, which prevents most emoji from being split into invalid surrogate halves. Some visible symbols use several code points, so a displayed grapheme can still count as more than one character.
The tool does not understand semantic sections, token budgets for a particular AI model, sentences, Markdown blocks, or attachment limits. It creates mechanical chunks and displays a visible separator only in the combined export; that separator is not part of the individual cards.
How to use it
- Paste the complete text you want to divide.
- Choose characters or words as the measurement unit.
- Enter the maximum size for each chunk.
- In character mode, choose whether to prefer word boundaries.
- Select Create chunks, inspect the cards, and copy or download the combined result.
If another service enforces a limit, leave room for titles, instructions, markup, or other text that will be added later. Confirm how that service defines characters, words, bytes, or tokens.
Benefits
- Supports character and word measurements
- Shows the measured size of every result
- Offers exact or word-preferred character boundaries
- Keeps the original order of the passage
- Adds clear separators to the combined export
- Performs all processing on the device
Using chunks responsibly
Read both sides of every boundary. A mechanical split can separate a heading from its explanation, a disclaimer from its subject, or a sentence from important context. Recombine or adjust chunks manually when meaning must travel with the excerpt. Do not treat chunking as anonymization; names and sensitive details remain unchanged.