Fix what is clearly broken.
Spelling, grammar, and punctuation corrections stay inside a narrow edit budget. Optional consensus passes filter one-off model choices.
Voice driftMINIMAL
Meaning protected. Edits bounded. Review transparent.
KEEP THE
MEANING.
MOVE THE WORDS.
Proofread the obvious. Smooth the awkward. Rebuild the structure. TextVeil shows its work and withholds edits that cross your quality limits.
Protected facts
No content logs
Rejected edits cost zero
SourceThe rollout was delayed due to several unexpected implementation issues that were encountered by the team.
Bounded editThe team delayed the rollout after running into unexpected implementation issues.
Choose the edit boundary
THREE LEVELS.
ONE SOURCE OF TRUTH.
You decide how far the model may move. Every route checks names, numbers, citations, URLs, code, and the terms you lock.
Make awkward prose read cleanly.
Paragraph-level edits improve cadence and clarity while preserving claims, sequence, voice, and approximate length.
Human reviewEXPECTED
Rebuild from the claim ledger.
TextVeil extracts the source claims first, then drafts a new structure. The result always stays visibly marked for meaning review.
Human reviewREQUIRED
01 / PARSE
Map the source
Claims, names, numbers, links, citations, code, and locked terms become the edit boundary.
02 / EDIT
Generate once
The selected model receives a mode-specific brief. Detector labels never enter the prompt.
03 / CHECK
Measure drift
Edit size, source coverage, protected facts, length, and optional semantic checks are evaluated.
04 / RELEASE
Return or withhold
Usable edits include review flags. Rejected and unchanged outputs consume no credits.
BEFORE & AFTER
A bounded edit can remove stiff wording without changing the point.
textveil --mode naturalize --lock-terms "AI, detection"
SOURCE DRAFT
The product launch was delayed due to several unexpected implementation issues that were encountered by the engineering team. The team is currently working to resolve these issues before a new launch date is announced.
AI-ASSISTED EDIT — REVIEW REQUIRED
The engineering team delayed the product launch after encountering unexpected implementation issues. It is resolving them now and will announce a new launch date when the work is complete.
Evidence before adjectives
THE NULL RESULT
STAYS ON THE PAGE.
Seven API models and 306 offline candidates narrowed the field on quality, cost, and latency. A staged Pangram 4.0 diagnostic then found no detector change. We publish both results.
OFFLINE SCREEN
306
Latest candidates across seven model arms, multiple modes, and mixed provenance controls.
PAIRED DIAGNOSTIC
0.0
Mean Pangram nonhuman-score change across all three finalists on five registered AI-fiction sources.
PRODUCT DECISION
QUALITY
No model was promoted on detector performance. TextVeil remains an editor, not a detector guarantee.
OPEN RESEARCH ARCHIVE
DETECTOR CLAIM: NOT ESTABLISHED
The actual product
A WRITING SURFACE,
NOT A SLOT MACHINE.
Bring a draft, choose the boundary, and compare the result with the source. Credit cost and review status stay visible throughout.
OPEN THE LIVE EDITOR
WHAT WE ACTUALLY DO
Not another "AI rewriter" that turns your thesis into a TikTok caption. This is surgical.
QUALITY GATES
Unsafe edits are withheld
Every candidate is checked for edit size, length drift, source coverage, names, numbers, protected spans, and story drift before it reaches you.
MODEL CHOICE
Three tested API options
Start with Qwen for the pilot quality/cost balance, choose DeepSeek for lower cost, or use Claude as the closed-model alternative.
TEXT TYPES
Fiction, Email, News, Science
Presets for novels, biographies, school books, casual notes, email, speeches, news, business, technical docs, legal copy, marketing, and scientific writing. Each type uses a matching syntax bank so a lab note does not wake up on a dusty road.
TERMINOLOGY LOCK
Code. URLs. Citations. Brand Names.
Lock the terms that must survive unchanged. The quality gate compares protected values and withholds candidates that alter them.
MCP / AGENT READY
Separate Keys For Agents
Pro and Agency users can issue named, revocable API and MCP credentials. Access is checked against the current subscription on every request.
REQUEST LIMIT
Up to 50,000 characters at a time.
The beta enforces a 50,000-character request limit. Review and process larger documents in meaningful sections.
THE ARCHITECT BEHIND THE CHAOS
Chad Devons
Vercel #14 Stripe Fraud Ops Google AI Platform Palantir FDE
I built this because I was spending 6–9 hours a week manually injecting domain expertise into AI agent context windows. The bottleneck was never the model. It was the knowledge. So I stopped being a copy-paste API and built the infrastructure layer instead. 5,900+ skills shipped. 37 domains covered. Still drinks cold coffee at 3 AM. Still ships on Fridays.
READ THE FULL LOREPRICING (BECAUSE FREE IS A LIE)
Monthly editing credits. One credit covers up to 100 source words; actual requests are capped at 50,000 characters.
TRIAL
FREE
- 1,000 words (10 credits)
- Three editing modes
- Three finalist models
- Terminology Lock
- Web editor only
- No credit card required
MOST POPULAR
STANDARD
$10 / month
- 100 monthly credits
- Quality-gated AI editing
- All three editing modes
- Terminology Lock
- Up to 10,000 source words monthly
- Google account sign-in
- Web editor only
PRO
$40 / month
- 500 monthly credits
- Quality-gated AI editing
- All three editing modes
- Terminology Lock
- Up to 50,000 source words monthly
- 2 named, revocable agent / MCP keys
AGENCY
$140 / month
- 2,000 monthly credits
- Quality-gated AI editing
- All three editing modes
- Terminology Lock
- Up to 200,000 source words monthly
- 10 named, revocable agent / MCP keys
WHAT THE PILOT ACTUALLY FOUND
Seven model arms produced 306 latest offline candidates across proofreading, unchanged-text, mixed-provenance, and owner-authored AI controls.
— reproducible model sweepQwen3.8 27B, DeepSeek V4 Flash, and Claude Sonnet 5 formed the detector-blind quality, cost, and latency frontier.
— offline Pareto screenThe staged AI-source Pangram diagnostic was also a null result: all five sources and all 15 finalist assignments remained 100% AI, including nine unique edited outputs.
— Pangram 4.0 staged paired diagnostic, n=5 sourcesThe sealed evaluation remains closed until genuine false-positive controls, consented mixed drafts, and blinded human ratings are collected.
— published no-go readiness reportCHANGELOG
What shipped. When. No fluff.
Loading release notes…
NEWS
The cleanup crew. Four parts, sourced, not polite.
Checking for the latest published notes…
The same money funded the theft. Now it funds the cleanup.
Anthropic’s $1.5B book settlement, Menlo’s Pangram check, Claude’s invisible watermark. Same month. The writer pays every window.
• part 1 of 4
First you pay to prove you’re human
Then you pay to look human. Institutions bought the fear. The person at the keyboard bought the alibi.
• part 3
Claude branded your sentences. It wasn’t in the deal.
Invisible SynthID, no opt-out, private detector. They say it doesn’t change your rights under the terms.
• part 4
FAQ (BECAUSE YOU'RE GONNA ASK)
Does this guarantee a Pangram or other detector result?
No. Detector scores are probabilistic, and our current diagnostic was inconclusive. TextVeil optimizes writing quality and fidelity; detector labels do not drive generation or approval.
Will it break my code or URLs?
Code spans, URLs, citations, names, numbers, and locked terms are checked. A candidate that changes protected content is withheld, but you should still review every edit.
Why so unhinged?
Because corporate AI tools sound like they were generated by a committee of HR managers who've never touched a keyboard. We're not that. We're the guy who fixes it at 2 AM with a Red Bull and a terminal.
Do you store my text?
TextVeil does not write source or output text to its request logs. The selected model provider still processes the request; OpenRouter calls request zero data retention and deny provider data collection.
What if I run out of credits?
Active Standard, Pro, and Agency subscribers can buy a one-time 10K, 50K, or 200K boost from the account page, then resume editing when checkout completes.
Do purchased credits expire?
One-time boost credits stay in your balance until used. A current paid subscription is required to buy a pack.