Powerful Capabilities

The redactX software helps automate a broad range of redaction workflows. Its multi-modal capabilities cover documents, images, videos, and audio files. By incorporating advanced artificial intelligence (AI) technologies—including natural language processing, computer vision, and speech recognition—redactX delivers state-of-the-art accuracy and a productive user interface, all in a single package suitable for on-prem installation as well as on cloud environments.

Some of its features and capabilities are described below.

Redact Documents

  • The software incorporates sophisticated natural language processing for Named Entity Recognition, enabling agencies to redact concepts (dates, locations, names, organizations, etc.). Over 95 standard entities are pre-configured. An included “zero-shot” entity recognition capability allows users to create custom entity definitions without system retraining.

  • Built-in advanced text analytics handle variations and imperfections such as typos or missing characters.

  • State-of-the-art OCR functionality handles many different situations.

  • With a few clicks of the mouse users create redaction criteria templates that are reusable, can be applied to process batches of documents, and can support automated, recurring scheduled tasks.

  • The software can recommend FOIA exemption codes that might apply to a redacted item

  • Forms data is intelligently extracted and used to expand the level of specificity and automation when redacting batches of complex documents.

  • The software incorporates significant workflow functionality, including:

    • Quick drill down of redacted items based on confidence level

    • One click corrections across documents

    • In a multi-user environment redaction jobs can be shared for review/approval.

    • Subsequent redaction revisions to a document are saved and can be compared.

    • Reports can be generated detailing all that was done to achieve a final redacted document.

    • Users can set up redaction tasks and schedule them for different types of documents using applicable redaction templates.

Redact Text on the Image/Video

redactX can detect and recognize text in over 90 different languages. The text can be rotated in either the <X,Y> or the <Z> plane. You can set redactX to detect and blur all text strings, or only some that you specify. Partial text string matches can also be anonymized. You have complete control.

Redact Faces

The software is able to redact people’s faces/heads at any angle in the “X, Y, or Z plane”. It works well in a broad range of lighting conditions, regardless of head sizes and quality of the input video. Partially occluded faces, for example from wearing surgical masks or Covid-19 style masks, are also detected and redacted. You can choose to redact all faces in a video or only those of specific individuals. You can define and maintain a library of faces that you want to blur in multiple videos. When it comes to anonymizing videos, demanding customers choose redactX as the best-of-breed tool in the industry.

Redact Objects

redactX can detect and anonymize a multitude of objects. The software includes a built-in library of objects. Type a few characters naming the object and the software will show you what is available to choose from. This built-in library is updated periodically with new content. redactX recognizes that sometimes users need to anonymize content that is atypical and not included in its libraries. A unique feature of redactX is that users can give the software a few examples of what they wish to blur, give those examples a name, and subsequently redact matching items in the video that have been assigned that name.

Redact Speech

Advanced multilingual speech recognition enables targeted word and entity redaction from audio files and video soundtracks. The system automatically generates transcripts for review and supports translation capabilities, allowing users to redact words across multiple languages within the same audio track. To redact spoken words a user simply provides a list of words to be redacted (typed in an input field or imported from a text document), specifies a match strength, chooses whether to bleep or mute the sound, and the software does the rest. Entity-level redaction of spoken words is also enabled. (e.x. “redact all names, locations and dates from the audio”)

Flexible Implementations

Based on customer requirements and preferences, redactX can be deployed on premises and in the cloud. On-prem installation guarantees that the complete process is under control of the user and file assets never leave their premises. Deployment is a simple matter of launching a software installer and configuring the server to do as you specify. The hardware configuration is typically a server with one or two Intel CPUs, multicore, and one or more NVIDIA GPU accelerator cards. While redactX will run on any modern computer, utilizing more cores, more processors, and more GPUs, will improve the speed of processing. Both Windows and Linux versions of the software are available. On-prem redactX is available on a conveniently priced yearly subscription business model.

For customers that prefer to run software on the cloud, redactX offers AWS Amazon Machine Images (ami’s), pre-configured and ready to go. These versions can be launched on a range of instance types, in all the international and domestic regions supported by AWS. Among these is the AWS GovCloud, which is a FedRamp certified cloud platform for US government agencies.

Launching an AWS instance can get you going in minutes, and you pay for usage by the hour. One and two year subscriptions are available at lower prices.

Administrator Mode

In an administrator capacity, a user can configure the server, manage users, set up recurring tasks, and create and manage databases.

A unique feature of the redactX software is that it gives users the ability to create and save databases of records that the software uses to carry out redactions. This capability enables advanced functionality such as controlling what is redacted in the video. It also enables making corrections to redacted video outputs.
An administrator can setup tasks, such as automatically monitoring the contents of a drive or an s3 bucket and redact new files according to rules in user-defined templates. An administrator can also network multiple instances of redactX to distribute the workloads across multiple servers.

A REST based Web Services API is also available for customers that wish to integrate the functionality of redactX in their own applications and workflows.