Tokenise prompt
Say what you need. Protect what matters. Set your budget.
Paste any prompt — research, academic, creative, coding, document analysis or an agent instruction. Scholar Speech rewrites it to use fewer tokens while keeping every objective, constraint, output requirement and rule you gave it.
Your prompt
Estimated tokens: 0 · 0 characters
Target model
Counts adapt to the model family. No provider tokenizer is available here, so all counts are shown as estimates rather than exact figures.
Optimisation mode
Removes repetition and filler, consolidates and restructures. Meaning kept.
Token budget
Protected instructions
Anything you protect is kept intact, even under maximum compression. If protection makes your budget impossible, protection wins and we tell you why.
Prompt check
No obvious instruction conflicts found. This is a simple check, not a guarantee.
Target: 220 tokens (automatic)