The seven tools
Re-Doc
Re-Doc finds personal and health information in native PDF, DOCX and TXT files and replaces it with realistic, consistent synthetic values, keeping the original layout. Scanned pages and images get permanent black-box redaction. Every result can be reviewed and edited before download.
- Good fit: documents that people still need to read, search or analyse after de-identification, such as clinical reports, case files, contracts and AI training sets.
- Deployment: web app and REST API on paid plans; on-premise deployment on request.
- Limits to know: it does not process video or audio, and synthetic replacement needs a text layer, so scans are redacted rather than replaced. Re-Doc does not hold SOC 2 or ISO 27001 certification yet; work toward security certification is in progress.
Redactable
Redactable is a document redaction platform. Its pricing page lists AI-powered redaction including signatures, logos and images, OCR for scanned documents, redaction certificates and audit logs, and metadata removal.
- File types: PDF, TIFF, PNG and JPG, up to 500 MB per file.
- Pricing: Starter at $19 or $23 a month for 15 documents, Pro Plus at $79 or $99 a month for 150 documents, Enterprise on request. The free tier covers 3 documents. API access is listed for high-volume users.
- Security: the site states SOC 2 Type 2 and HIPAA compliance.
- Good fit: teams that need visible, permanent redaction of PDFs and images with simple monthly pricing.
CaseGuard
CaseGuard Studio redacts PII across documents (PDF, PST, Excel and Word), body-worn camera, CCTV and dash-cam video, audio, and images.
- Deployment: "CaseGuard runs locally on your machines, including air-gapped networks without internet."
- Customers: the site lists law enforcement, government, legal, healthcare, education, insurance, call centers and more, and names FOIA among the standards it supports.
- Pricing: not listed on the homepage; the site says every plan includes unlimited redactions and file uploads.
- Good fit: records and law-enforcement teams that handle video and audio as well as documents.
Limina (formerly Private AI)
Limina detects and removes personal data from "text, images, audio, and documents", with 50+ entity types across PII, PHI and PCI, in 52 languages.
- Deployment: "Container runs in your VPC or on-prem."
- Options: redact, pseudonymize, generalize or generate synthetic PII. The page does not say whether document layout is kept.
- Trust signals on the site: an ISO 27001 badge and independent validation by Armilla AI. Named customers include Boehringer Ingelheim, Zurich Insurance, MUFG Bank and Providence.
- Good fit: developers and data teams that need a detection API across many languages, run in their own environment.
Tonic Textual
Tonic Textual is described as "an unstructured data redaction and synthesis solution". It can redact or synthesize sensitive entities consistently, and lists file types including PDF, DOCX, PPTX, PNG, JPEG, CSV and audio.
- Deployment: cloud-hosted, in your own cloud through S3, GCS or Azure Blob Storage, or self-hosted with Kubernetes or Docker. Also on the AWS, Google Cloud and Snowflake marketplaces.
- Layout: the FAQ says it "preserves context, intent, and structure"; the page does not say whether the visual layout of a document is kept.
- Good fit: teams preparing unstructured data for AI training, RAG and testing.
John Snow Labs
John Snow Labs' medical de-identification covers clinical text, text and scanned PDFs, DICOM images (burned-in PHI and metadata), structured data and FHIR.
- Options: replace names with realistic fakes, mask fields, shift dates while keeping intervals, generalize locations, and tokenize values for linkage.
- Deployment: on-premise, cloud or air-gapped; runs inside Databricks, Snowflake, Azure, AWS or your own datacenter.
- Vendor-reported results: 0.98 micro F1 on the 2014 i2b2 test set, and two billion patient notes de-identified at Providence. These are the vendor's figures.
- Good fit: healthcare and pharma data teams working at very large scale inside a data platform.
Microsoft Presidio
Presidio is an open-source framework (MIT licence) "for detecting, redacting, masking, and anonymizing sensitive data". It covers text, images including DICOM, and structured data.
- What you get: building blocks, not a finished product. You write the pipeline, the review step and any document handling yourself. The project page does not mention keeping PDF or DOCX layout.
- Its own warning: "there is no guarantee that Presidio will find all sensitive information."
- Good fit: engineering teams that want a free base to build their own tool.
