Reading time: 12 min
Amped Replay’s new Assisted Redaction speeds up video redactions by automatically detecting people, heads, vehicles, and license plates. This article shows how to install the optional ML component, prepare footage for best results, review detections, refine hide and label settings, and export a fully redacted, defensible MP4 and report.
Hello everybody and happy new year to you all! We kick off 2026 with a dedicated article on the brand new Assisted Redaction feature, which was released with the latest Amped Replay version near the end of last year. With this article, we delve deep into the usage of the new filter so that no stone is left uncovered! Join us if you want to learn some useful hints and tips and start your journey into assisted video redactions right away!
Key Takeaways
- Assisted Redaction in Amped Replay uses machine learning to detect persons (full body), human heads, vehicles, and license plates, creating editable bounding boxes and optional text labels for faster video redaction.
- Installation is optional by design: download the build with Assisted Redaction only if your organization permits ML tooling; otherwise install the build without it (and you will not see the Assisted Red. button).
- Performance is GPU-accelerated: Assisted Redaction leverages the GPU when available (CPU fallback), so a stronger GPU can materially reduce analysis time on long clips and busy scenes.
- Prepare your footage for speed and accuracy: select only the relevant range in the Play tab and apply any helpful Enhance filters (e.g., contrast/light adjustments) before running detection.
- Use the Sort Assisted Video Redaction panel to manage results at scale: visually identify objects via thumbnails, filter the list to the current frame, merge duplicate detections, redo tracking, and remove false positives.
- After verification, click Apply All to automatically apply hide filters to all detected objects; if the detector misses some frames, the workflow can apply automatic interpolation (keyframing) to maintain continuity.
- You can standardize output quickly by customizing hide type/strength/shape and then applying those properties across all Hide and/or all Text labels in one action.
- Human review is mandatory: play/scrub the redacted clip and correct any misses or dropouts (e.g., occlusion or appearance changes) before export, because the user remains responsible for complete coverage.
- For delivery, export “Export Processed Video as MP4” to produce a fully redacted video, and generate a PDF report documenting redactions and noting that ML assistance was used.