Review & deliver
Smart Proofread before you deliver
After twenty pages, the eye stops catching things. Smart Proofread re-reads the whole translated document against your glossary, your styleguide, and the source — and flags every drift, every missed term, every register slip. Run on a selection or the whole document.
Free to begin · No card required
Featured on
Even good translations drift. After twenty pages the eye stops catching the terminology slip; after fifty pages, the register starts to wobble. Smart Proofread is the dedicated QA pass — a second read of the translated document against the source, against the glossary, against the styleguide — that flags every drift and proposes a fix. It’s the most-skipped step in professional translation because it’s tedious by hand. Smart Proofread automates the work so the QA pass becomes the default before delivery instead of the thing you only do when you remember.
How does a second pair of eyes read the whole document?
Manual QA is the most-skipped step in translation. Smart Proofread makes it the default — a deterministic pass that catches the kinds of misses you can’t catch by re-reading.
Reads against your glossary
Every glossary term in the source is checked against the target. Missed mappings, inconsistent renderings, accidental synonyms — all flagged.
Checks the styleguide
Register slips, sentences that broke the tone, paragraphs that lost the voice. Smart Proofread compares the translation against the styleguide rules.
Surfaces sentence-level issues
Drift, mistranslation, omissions, additions, ambiguity. Each issue ties back to the sentence, with a suggested fix you can accept, edit, or reject.
Approve in bulk or one by one
Either approve every suggested fix in one click, or walk through the review page sentence by sentence — same UX as the editor.
What does it catch that re-reading won’t?
Terminology drift
The product term that started as "tray" and became "panel" by page 40. Caught.
Register slip
The chapter that started formal and slowly turned casual. Caught.
Number, date, and unit errors
Localized numerals, date formats, currency, units. Caught and fixed.
~1,000 calls measured
We logged the reasoning on roughly a thousand real translation calls. On plain text the pricier Pro model thinks about half as much as Flash. The harder the sentence, the more any model reasons, whatever its tier. So we spend Pro only where it earns its cost: the passages a cheaper model would fumble.
What we measured— Don’t take our word for it —
The editor, in miniature
A working slice of the real thing — Literess, glossary, styleguide, workflows, and the translation memory are all live. Click around.
Here to help you translate
A key turned in the lock and the door swung open.
У замку повернувся ключ, і двері розчахнулися.
The knock came just before midnight.
Стукіт пролунав перед самою північчю.
FAQ
- Yes — but only for the blocks it actually fixes. Blocks that already read well are skipped entirely: no model call, no words charged. Fixed blocks bill at your document’s mode rate.
- Yes — select any range and run Smart Proofread on just that. Useful for chapter-by-chapter book work.
- Hand-edits are treated as locked-in choices. Smart Proofread flags drift relative to them, not against them.
- Smart Proofread reads the text. Vision Proofread looks at the rendered page like a reader would — catching layout, visual, and rendered-output issues that text-only review misses.
- It’s the layer that makes that workflow practical. Raw machine translation gives you a draft; Smart Proofread is the automated translation review pass that re-reads it for drift before a human ever looks — so your human review is spent on judgment calls, not hunting for a glossary term that slipped on page 40. AI-powered MT with human review, with the tedious first sweep done for you.
- Glossary compliance (missed terms, inconsistent renderings, accidental synonyms), styleguide compliance (register slips, tone drift, sentence-rhythm changes), terminology consistency across the document, number/date/unit handling, and omissions or additions relative to the source.
- Usually 30–60 seconds per 1,000 words. A 50-page document runs in under five minutes. The review UI shows issues in real time as they’re found, so you can start working through them before the pass finishes.
- Yes — select any range and run on just that section. Useful for chapter-by-chapter book work or when you only changed part of a document.
- Sometimes — especially when your styleguide deliberately departs from convention. Each flag has a confidence indicator; low-confidence flags are easy to dismiss. The pass gets better as you reject false positives, because they feed back into the styleguide.
- It catches what a tired human editor misses (terminology drift, register slips) but not what a fresh human editor catches (cultural fit, deeper meaning, hard-to-articulate "feel"). The combination is what most professional teams use — Smart Proofread for the consistency pass, human review for the judgment pass.
Keep reading around this capability
- FeatureA glossary the AI actually follows→Character names, product terms, branded phrases, client-specific vocabulary. Build a glossary once — manually, by upload, or seeded from a…
- Use caseTranslate the next client document→Per-client glossaries that travel across projects. Per-client styleguides that hold tone. Sentence-level review for the lines that matter.…
- Use caseTranslate policy conservatively→Legal and policy translation is the opposite of creative. Preserve terms exactly. Don’t paraphrase. Flag uncertain clauses for human…
Deliver every document with a second pass behind it
Start translatingFree to begin · No card required