Reveal Trace
Encode, decode, and expose obfuscated text.
Result
Your result appears here.
What Reveal Trace does
Reveal Trace is a free online tool that encodes plain text into obfuscated forms and decodes obfuscated text back to plain text. Paste something encoded to reveal it, or type plain text to disguise it, across 67+ methods. It exists to show how attackers hide prompt-injection payloads from LLM filters, and how defenders can decode and detect them.
Methods it covers
- EncodingsThe model decodes it; the filter sees no words.
- CiphersLetter shift or substitution.
- Look-alikesSwap letters for confusable characters.
- InvisibleHidden or zero-width characters.
- Fancy fontsFancy Unicode fonts that still read the same.
- Other alphabetsBraille, runes, kana, hieroglyphs.
- WordplayHuman-readable mangling of the words.
Frequently asked questions
- How do I decode obfuscated or encoded text?
- Paste the text into Reveal Trace and switch to Decode. It detects the likely method (Base64, hex, ROT13, homoglyphs, zero-width characters and more) and shows the decoded plain text.
- What is a homoglyph generator?
- Homoglyphs are look-alike characters from other alphabets (for example Cyrillic 'а' instead of Latin 'a'). Reveal Trace's homoglyph method swaps letters for confusable characters that read the same but have different code points, which is how text slips past exact-match keyword filters.
- How do I add invisible or zero-width characters to text?
- Use the Invisible text / zero-width methods in Encode mode. They insert characters that render as nothing but are still read by many language models — useful for demonstrating hidden-text injection.
- How is text obfuscation used to jailbreak LLMs?
- Encodings and look-alike characters hide banned words from input and output filters while the model still understands the content. Reveal Trace lets you reproduce these evasions and decode them, pairing with the Prompt Injection Cheat Sheet.
- Is Reveal Trace free?
- Yes. Reveal Trace is a free tool on PromptTrace, a hands-on AI-security training platform. No sign-up is required to encode or decode text.
Related
Pair Reveal Trace with the Prompt Injection Cheat Sheet (133 techniques & evasions) and practice on real models in the labs. Read up on homoglyph attacks and zero-width & Unicode steganography.