# Phravv > Phravv is a high-fidelity paraphrase tool at https://www.phravv.com. Paste text. Get a high-fidelity paraphrase. Same meaning. New wording. Phravv rewrites pasted text while keeping facts, names, numbers, dates, quotations, URLs, heading lines, and fenced code. Light, Balanced, and Aggressive decide how far the sentences move. They are not stealth settings. A thorough paraphrase can degrade Anthropic Claude’s statistical token watermark (a SynthID-Text-style pattern in token choice, not hidden characters). That is not a guarantee. Short text is inconclusive. Phravv is not a detector and never rewrites with Claude. ## Pages - [Phravv](https://www.phravv.com): the rewriter - [FAQ](https://www.phravv.com/faq): Claude’s text watermark, and what a paraphrase does - [Light vs Balanced vs Aggressive](https://www.phravv.com/guides/light-balanced-aggressive): How Phravv’s Light, Balanced, and Aggressive strengths work. They decide how far sentences move while keeping the meaning. Craft controls, not stealth settings. - [Paraphrase vs humanizer](https://www.phravv.com/guides/paraphrase-vs-humanizer): What Phravv is and is not. A high-fidelity paraphrase keeps meaning and changes wording. It is not a humanizer, a detector, or a way around checkers. - [What Claude watermark claims can and cannot do](https://www.phravv.com/guides/claude-text-watermark): An honest note on Claude’s statistical text watermark and what a paraphrase can and cannot do. Cited from Anthropic. Not a guarantee, and not a detector-bypass guide. ## What Phravv is not Not a detector. Not a humanizer. No claim of undetectable text, detector bypass, or 100% watermark removal. ## Sources - [How Claude’s text watermark works](https://www.anthropic.com/news/claude-text-watermark) - [How Claude marks AI-generated content](https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content) ## FAQ ### What is Phravv? A paraphrase tool at phravv.com. Paste a draft, pick a strength, and get a second pass that keeps the meaning and changes the wording. The facts stay. The sentences move. ### How do Light, Balanced, and Aggressive differ? They decide how far the sentences move. Light keeps sentence order and leans on synonyms. Balanced rephrases most sentences and keeps the argument order. Aggressive is allowed new sentence boundaries and voice, with the same facts. They are craft controls, not stealth settings. ### What does Phravv leave alone? Numbers, names, dates, quotations, URLs, heading lines, and fenced code. It is a constrained rewriter, not a new draft from a prompt. ### Do you keep what I paste? The text is sent only to produce the rewrite. We don’t keep a server-side history, and we don’t sell what you paste. Recent pairs may stay on this device in your browser until they’re overwritten or you clear site data. ### How long a draft can I paste? About 4,000 characters, enough for a short document. Longer pieces work in sections. ### What is Claude’s text watermark? When Claude writes, it leaves a statistical pattern in the words it chooses. Anthropic’s method is a version of Google DeepMind’s SynthID-Text. Nothing is inserted into the file, and there are no hidden characters. You can’t see it on the page. Anthropic described it on August 14, 2026. Sources: https://www.anthropic.com/news/claude-text-watermark https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content ### What does a paraphrase do to it? A thorough rewrite can weaken the signal. Anthropic’s own note is that light edits probably won’t clear it, while replacing the words can. We don’t promise that, and we don’t call the result undetectable. Short passages don’t give you much to go on. Dense facts and code have fewer places for a mark to live in the first place. If a mark is found, it only means Claude was likely involved at some point. It doesn’t say who wrote the piece. Sources: https://www.anthropic.com/news/claude-text-watermark ### Why doesn’t Phravv rewrite with Claude? If Claude rewrites the text, it can put the same kind of mark back in. Phravv never calls Anthropic models. ### Do you detect the mark? No. No badge, no score, and we aren’t building a detector. If Anthropic opens a public detection API, we might use it behind the scenes and add a sentence here. The rewriter won’t depend on it. We don’t sell a way around detectors, and we don’t pitch this at students trying to beat school checkers. ## Guides ### Light vs Balanced vs Aggressive Phravv’s three strengths are craft controls. They decide how far the sentences move. The facts stay. The meaning stays. They are not stealth settings, intensity knobs, or a way to hide a draft. #### What the strengths share Every strength is a constrained paraphrase. Numbers, names, dates, quotations, URLs, heading lines, and fenced code stay. You are not prompting a new document. You are asking for a second pass on the one you pasted. #### Light Light keeps sentence order and leans on synonyms and light clause reordering. Voice and contractions stay. Use it when the draft is almost ready and you want a quieter second pass — a sendable polish, not a recast. #### Balanced Balanced rephrases most sentences and keeps the argument order. The through-line of the piece stays put; the wording moves. On the current model this often sits close to Light. It is the default when you want a fuller paraphrase without new sentence boundaries. #### Aggressive Aggressive is allowed new sentence boundaries and voice, with the same facts. A heads-up stays a heads-up. It is not “make it weirder,” and it is not a policy memo. Check the register on casual mail — the facts should still be yours. #### How to choose Start with Light if you would send the original with a few wording nits. Use Balanced when the sentences need to move but the outline is right. Use Aggressive when the structure itself is stiff and you still need the same claims. If a pass goes too far, paste the output back as input and step down. #### What they are not They are not detector settings. They do not promise a cleaner mark, a human-sounding score, or an undetectable result. Farther movement is editorial, not covert. ### Paraphrase vs humanizer Phravv is a high-fidelity paraphrase tool. That is a different job from a “humanizer.” The distinction matters, because the two words get sold as if they were the same. #### What a paraphrase is A paraphrase keeps the meaning and changes the wording. Facts, names, numbers, and quotations stay. Sentence shape can move. The author still owns the claims. That is editorial work: a second pass on a draft you already wrote or assembled. #### What “humanizer” usually means Tools that call themselves humanizers usually pitch a different outcome: text that will pass as if no model was involved. The promise is often a score, a bypass, or “undetectable” prose. That is not editing. It is a claim about checkers. #### What Phravv is Phravv rewrites pasted text. You pick Light, Balanced, or Aggressive — how far the sentences move — and get a second pass. The homepage is the rewriter. Quality of the rewrite is the point. #### What Phravv is not Not a humanizer. Not a detector. Not a way around school or workplace checkers. We do not sell undetectable text, detector bypass, or 100% watermark removal. We do not pitch this at students trying to beat those tools. #### Why we draw the line A meaning-preserving rewrite can change how a passage reads, and it can weaken some statistical marks. That is a side effect of changing the words, not a product promise. If you need a cleaner argument or a sendable note, paraphrase. If you need to hide authorship from a checker, this is the wrong tool. ### What Claude watermark claims can and cannot do Claude can leave a statistical pattern in the words it chooses. A thorough paraphrase can weaken that signal. That is not a guarantee, and this page is not a guide to beating detectors. #### What the watermark is When Claude writes, it leaves a statistical pattern in token choice. Anthropic’s method is a version of Google DeepMind’s SynthID-Text. Nothing is inserted into the file. There are no hidden characters. You cannot see it on the page. Anthropic described it on August 14, 2026. Sources: https://www.anthropic.com/news/claude-text-watermark https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content #### What a paraphrase can do A thorough rewrite can weaken the signal. Anthropic’s own note is that light edits probably will not clear it, while replacing the words can. Dense facts and code have fewer places for a mark to live. If a mark is later found, it only means Claude was likely involved at some point. It does not say who wrote the piece. Sources: https://www.anthropic.com/news/claude-text-watermark #### What those claims cannot do We do not promise the mark is gone. We do not call the result undetectable. Short passages do not give you much to go on. A light synonym pass is not a thorough rewrite. Mixing a few original sentences back in can keep a signal that a full pass would have moved. This is not a detector-bypass guide. We do not publish settings, tricks, or success rates for clearing a mark. #### What Phravv does not do Phravv does not detect the mark. No badge, no score, and we are not building a detector. If Anthropic opens a public detection API, we might use it behind the scenes and add a sentence to the FAQ. The rewriter will not depend on it. Phravv never rewrites with Claude. If Claude rewrites the text, it can put the same kind of mark back in. #### How to read this Use the rewriter because you want a second pass that keeps the meaning. Read Anthropic’s own notes if you care about the mark. Treat any “guaranteed removal” claim — ours or anyone else’s — as marketing, not evidence.