Paste text or drop a .txt file. Works with Granola, Zoom, Otter, or any text source.
📄Drop a .txt file hereor click to browse
0 characters
How it works: Load a transcript → smart scanner detects patterns → click to confirm or add redactions → optionally run a Claude AI deep scan → export as clean text, FBI redacted, or code-named with reversible key file.
Select & redact
Click any word to redact it. Click adjacent words to build phrases. Dotted lines = suggestions. Solid = confirmed.
People
0
Orgs
0
Projects
0
Sensitive
0
Instances
0
0 / 0 suggestions confirmed
People
Orgs
Projects
Sensitive
Suggested
AI scan
🔍
Scanning...
Dictionary 0
Click words in the transcript to add them.
Persistence
Export
Three formats. Clean text for AI. FBI redaction for sharing. Code-named with reversible key file for round-trip editing.
📝 Clean Text
[TYPE] tags replace redacted terms. Safe for AI.
🔒 Redacted Document
Black-bar redaction. Printable. Unrecoverable.
🔑 Code Named
Codename replacements. Reversible with key file.
Restore
Paste a code-named document + key file to swap codenames back to originals. For internal use only.
Code-named document
Key file
Restored Document
👤 Person
🏢 Org
📁 Project
🔴 Sensitive
🛡 Trust & Safety
Blackout is designed with a security-first architecture. Here's exactly how your data is handled.
Local Processing — Always On
Every transcript you load is processed entirely in your browser using JavaScript. The text never touches a server. All redaction, tokenization, codename generation, and export happens in memory on your device. When you close the tab, the transcript is gone — nothing is stored except your dictionary (in localStorage, which you control).
You can verify this yourself: right-click → View Source. There are zero fetch() calls, zero XMLHttpRequest calls, zero WebSocket connections in the core tool. The only network request the tool can make is the optional AI deep scan, which you must explicitly trigger.
AI Deep Scan — Optional, Explicit, Transparent
If you configure an API endpoint in Settings and click "Deep Scan with Claude," your transcript text is sent to the Claude API via your configured proxy endpoint. This is the only moment data leaves your machine, and only when you click the button.
What Claude sees: Your transcript text, plus a system prompt asking it to identify entities. Nothing else.
What Claude does with it: Anthropic's API has a zero data retention policy for API calls. Your transcript is not stored, not logged, not used for training, and not accessible to any human at Anthropic. It is processed and immediately discarded.
When it's offline: The button is grayed out. No request is attempted. If a request fails (no network, wrong endpoint), the tool catches the error locally and shows you the error. No data is cached or queued for retry.
The Proxy Endpoint
When you click "Deep Scan," your transcript is sent to a secure API endpoint that forwards it to Claude. Your Anthropic API key is stored server-side — it never appears in the browser, in the page source, or in any response. You authenticate with your email, and the server handles the rest.
If you don't have a Pro subscription, the AI scan is simply unavailable. The rest of the tool works identically.
What's Stored Locally
localStorage (browser only, you control):
• Your redaction dictionary (terms + types) — so you don't re-enter names every session • Your API endpoint URL and login token — so you stay logged in • Nothing else. No transcripts. No exports. No history.
Clear your browser data to remove everything. Or click "Clear" in the sidebar.
⚙ Settings
Connect to Blackout Pro for AI-powered deep scanning. The local scanner works without an account.
Connection
API Endpoint
Your Blackout API URL.
Account
Email
Enter the email you used to subscribe, or sign up free to try the local tools.
Free plan: Full local scanner, click-to-redact, all export formats, dictionary persistence. Pro plan ($15/mo): Everything free, plus AI deep scan powered by Claude. Subscribe here →