// compress
Cut your token count without losing meaning.
Paste any prompt or document. Filler phrases are removed, redundancy stripped, whitespace cleaned — fewer tokens, same meaning. Runs entirely in your browser.
Compressed output appears here
Paste text on the left to start
// how it works
Fewer tokens. Same meaning.
English is full of filler — phrases that add words without adding meaning. “In order to” means “to.” “Due to the fact that” means “because.” “At this point in time” means “now.” When you send text to an LLM, every one of those extra words costs you tokens.
The compressor applies a library of pattern replacements — longest-to-shortest, case-preserving — then cleans up whitespace and redundant structure. It runs entirely in your browser: no API calls, no data sent anywhere.
Typical savings range from 8–25% on real-world documents. Formal writing, legal text, and corporate prose tend to compress the most.
in order to achieve the goal
to achieve the goal
−4 tokdue to the fact that it failed
because it failed
−4 tokat this point in time
now
−5 toka large number of users
many users
−4 tokis able to process
can process
−2 tok// compression levels
Choose your compression level.
Light
5–10% reductionOnly the most obvious filler phrases — "in order to" → "to", "due to the fact that" → "because". Safe for any context. Output reads identically to the input.
Best for: Legal docs, formal reports, academic text where you need to preserve tone exactly.
Balanced
10–20% reductionFull filler library + discourse markers + pleonasms ("end result" → "result"). Also cleans trailing whitespace and excess blank lines. Recommended for most use cases.
Best for: Prompts, system messages, documentation, any text you paste into an LLM chat window.
Aggressive
15–30% reductionAll above + nominalisation reversal ("make a decision" → "decide") + discourse markers removed ("in conclusion", "to summarize"). May slightly change tone but not meaning.
Best for: Large context windows, RAG chunks, batch embeddings where token cost matters most.
// faq
Frequently asked questions
Does compression change the meaning?
At Light and Balanced levels, no — the replacements are semantically equivalent. 'In order to' and 'to' are identical in meaning. At Aggressive, some phrasing changes (nominalisation reversal) that may slightly alter tone but not content.
How much does it actually save?
Typically 8–25% on real-world text, depending on how formal the writing is. Corporate prose, legal text, and academic writing compress the most. Conversational or already-tight text may show 3–8%.
Is my text sent anywhere?
No. Everything runs in your browser using plain JavaScript pattern matching. No API calls are made. Your text never leaves your device.
What kind of text compresses best?
Formal written English — corporate memos, legal documents, academic papers, product specs, email threads. Conversational writing and already-concise prose compress less because the filler is already absent.
How is this different from summarization?
Summarization removes content. This removes only the words that don't carry content — filler phrases, pleonasms, discourse markers. The compressed version contains all the same facts and logic as the original.
Should I review the output before using it?
Yes, always. The tool shows how many substitutions were applied. Read through the output — especially with Aggressive level — to confirm the phrasing still reads naturally in your context.