Guides for teams integrating image, video and text redaction: how to accept results, choose a style, stay within PIPL, integrate the API and control output quality. Based on our own evaluation and integration work; updated from time to time.
Articles in this section are written in Chinese; the title and abstract are in English.
How to integrate the ORedact image and video redaction API: sign up for the console and create an API key, upload a file, submit a task, poll its status and fetch the result, verify webhook callbacks with the HMAC signature, handle every error code, and understand the trial allowance and billing rules, with request and response examples for each endpoint.
Read the articleThe full path to integrating the image and video redaction API: the 14-day trial allowance, sample validation and acceptance criteria, the three-step asynchronous task flow, callbacks and billing rules, retention and concurrency agreements, and a go-live checklist to prepare before integration starts.
Read the articleCompares the three redaction styles on reversibility, appearance and compliance. Blur and mosaic reduce information irreversibly; solid fill replaces pixels outright. Gives recommended styles for faces, plates, ID numbers and on-screen text, plus the default strength and padding rules used in production.
Read the articleExplains how China's Personal Information Protection Law treats facial images as sensitive personal information, the limits on image capture in public places, when redaction counts as anonymisation, when an impact assessment is required, and what to confirm about retention and training use before choosing a vendor.
Read the articleWhy redacted video files used to come out two to three times larger than the source, how re-encoding in pure quality mode inflates bitrate, what capping at the source bitrate and matching the source specification changed in measured output, and how colour tags and preview proxies fit into the pipeline.
Read the articleA practical acceptance method for redacted video: define what counts as a miss, grade misses by severity, sample frames side by side instead of trusting aggregate counts, track a visible-frame rate, and keep a simple log that feeds misses back into regeneration.
Read the articleSign up, validate on your own footage for free, then agree on integration and pricing; for self-service, use Damaya.
