Most filler tools shuffle the same twenty stock sentences, so a five paragraph batch starts repeating itself by the third block. This one builds every sentence on the server from typed templates and word pools, then tracks what it already used. Pick a style, set the length, and take the text as plain paragraphs or HTML.
Nothing generated yet
Choose a style and a length on the left, then hit generate. Each request builds fresh text on the server.
Filler text has one job: hold the shape of real copy long enough for someone to judge a layout. It fails at that job the moment a reader spots the same sentence three times in one column.
Browser based filler tools ship a fixed array of sentences and pick from it at random. That approach caps out fast. Twenty sentences means twenty possible lines, and a batch of ten medium paragraphs pulls roughly sixty sentences, so repeats are guaranteed before you finish scrolling.
Here the sentences get assembled in PHP on each request, before they reach your screen. The engine holds sentence templates with typed slots, plus separate word pools for each slot type. A template like The {adj} {thing} matters more than {people} expect resolves against three pools and produces thousands of distinct lines on its own.
The counter under the style picker reports the real number of distinct sentence forms for the selected style, calculated from the template list and the pool sizes rather than guessed at. Plain English clears 2.7 million sentence shapes. That figure counts single sentences, so paragraph level collisions sit far below anything you would notice by eye.
| Style | Reads like | Use it for |
|---|---|---|
| Plain English | Short declarative sentences about pages, spacing, and readers. | Client mockups where stakeholders read the words and comment on them |
| Business | Rollouts, retainers, burn rate, sign-off, quarterly language. | Slide decks, report templates, SaaS dashboards, proposal layouts |
| Engineering | Queues, cold starts, p99 latency, on-call pages. | Developer docs, changelog mockups, status page designs, admin panels |
| Narrative | Harbours, lanterns, weather, a slower rhythm and longer clauses. | Book covers, blog card previews, fiction publishing themes |
| Lorem ipsum | Latin word salad with varied sentence lengths and commas. | Type specimens and any review where reading the words is a distraction |
The choice between Latin and English is a real trade, not a preference. Latin stops a client editing your placeholder copy in a review, because nobody argues with consectetur adipiscing. English filler does the opposite: it exposes bad line lengths, awkward widow lines, and hyphenation faults that Latin hides, since Latin word lengths do not match English word lengths. Use Latin in front of clients, English while you are still building.
Two output formats cover most of what people paste this into. Plain text separates paragraphs with a blank line, which is what a design tool, a text field, or a spreadsheet cell wants. The HTML option wraps each block in <p> tags, ready to drop straight into a template or a static page.
Copy buttons sit on each paragraph as well as on the batch, so you can take one block without regenerating. The TXT download names the file after the style you picked, which helps when you are testing several styles against the same layout and end up with a folder of them.
Where this falls short
The English styles are grammatical, not meaningful. Sentences pass a read-aloud test and then say nothing, which is the point of placeholder text and also its risk: filler that scans as real prose has a habit of surviving into production. Search a codebase for a stray word from your chosen style before launch. Two more limits worth knowing. Word pools are English only, so this gives you nothing for testing CJK line breaking or right-to-left layouts. And the batch cap is 25 paragraphs per request, since the response is built and returned in one pass rather than streamed.
Designers usually arrive with a character budget, not a paragraph count. Averaged over 120 generated paragraphs per setting, this engine produces about 30 words on short, 63 on medium, and 101 on long. Latin and English land within a word of each other at the same setting, since the lorem sentence lengths were tuned against the English ones. Switching style will not move your layout.
Work backwards from there. A 500 word hero block is roughly eight medium paragraphs. A card grid wants the short setting, because medium paragraphs overflow a card and hide the overflow behaviour you are trying to test. Averages are averages, so when a slot needs an exact count, paste the output into the word counter and trim.
For a single line rather than a block, the random sentence generator fits better, and the random word generator covers labels, tags, and placeholder names. Anyone who specifically wants the traditional Latin blocks with no style switching should use the dedicated lorem ipsum generator instead.
A request builds the paragraphs and returns them. Nothing is written to a database, no record ties a batch to you, and closing the tab ends it. Hit generate again and the engine composes new text from scratch rather than reading anything back.
The output is free to use in commercial work, client deliverables, and published templates. It is machine assembled text with no source material behind it, so there is no attribution to give and no licence to track.
Repeats, output formats, and using filler in real projects.
Not in normal use. Plain English alone has more than 2.7 million distinct sentence forms, and a paragraph chains four to eleven of them, so an identical paragraph is not something you will hit by generating. Individual sentences repeat sooner than that, mostly when you narrow the style and run long batches back to back.
Shipping the engine to the browser would mean shipping every template and word pool as a JavaScript payload, which slows the page load and caps how large the pools can grow. Keeping composition in PHP lets the word lists expand without adding a byte to page weight, and it keeps the templates out of view so the output does not read as an obvious pattern.
Yes. The text is assembled by an algorithm from generic word lists, not copied from any source, so there is no attribution requirement and no licence attached. Use it in client work, themes you sell, and public demos.
No. Every request composes new text, and nothing is stored, so a batch you liked is gone once you regenerate. Copy or download anything you plan to reuse before changing the settings.
Each hides a different problem. Latin keeps reviewers focused on layout, since there is nothing to read. English matches real word length distribution, so it surfaces bad line lengths, orphans, and hyphenation faults that Latin masks. Most teams use English during build and switch to Latin for client review.
The whole batch is composed and returned in one response rather than streamed, so a very large request would leave the page waiting with nothing on screen. Twenty five long paragraphs runs past 2,500 words, which covers a full page mockup. Generate twice if you need more.
