How to redact a region of video before use as legal evidence
- Step 1Load the evidence file locally — Drop the MP4, MOV, MKV, AVI or TS exhibit. It loads into an in-browser FFmpeg.wasm filesystem on your machine — no upload, no server copy. Process one file per run to keep each exhibit's handling auditable.
- Step 2Identify the exact frame and feature — Set Start (s) to the frame where the protected feature first appears; the picker renders that frame so you can position the box precisely over it (a bystander's face, a document, a screen).
- Step 3Draw the redaction rectangle — Drag the box over the feature. The live label reads
W × H @ (X,Y)in source pixels — note these values for your disclosure record. Resize with the SE corner; reposition by dragging inside. - Step 4Define the time window — Enter End (s) for when the feature leaves frame. Record Start and End. The redaction is active only across that window, so the rest of the exhibit is unaltered and remains usable as evidence.
- Step 5Set and record the blur strength — Choose Blur strength (5–50). Note that the effective
boxblurradius ismin(strength, min(width,height)/2)— for a small region, document that the radius was capped and enlarge the box if legibility remains. - Step 6Export, verify, and document — Process to H.264 MP4 and verify the feature is unreadable across the entire window. Record the full parameter set (
X, Y, W, H, Start, End, strength), the output codec (H.264 CRF 20), and the on-device process in your disclosure notes.
Parameters to record for a defensible redaction
Every value that defines the redaction, and where it appears, so it can be documented and reproduced.
| Parameter | Where it's shown | Maps to FFmpeg |
|---|---|---|
X, Y | Picker label @ (X,Y) + numeric fields | crop/overlay offset (source pixels) |
Width, Height | Picker label W × H + numeric fields | crop dimensions |
Start (s), End (s) | Numeric fields | enable='between(t,start,end)' |
Blur strength | Numeric field (5–50) | boxblur radius (capped at min(W,H)/2) |
| Output codec | Fixed by processor | libx264 -preset medium -crf 20 |
| Processing location | Browser / FFmpeg.wasm | No upload — on-device |
Evidence-handling properties
What the tool guarantees for disclosure work.
| Concern | Behaviour |
|---|---|
| Source ever uploaded? | No — processed in-browser on your device |
| Redaction reversible? | No — boxblur destroys the region's pixels |
| Audio altered? | No — stream-copied unless you mute separately |
| Metadata stripped? | No — use metadata-scrubber if disclosure requires it |
| Original file modified? | No — a new MP4 is produced; keep the original sealed |
| Reproducible from notes? | Yes — record the parameter set above |
Cookbook
Recipes for redacting exhibits for disclosure. Record every parameter for your notes.
Blur a bystander's face in CCTV for disclosure
A third party uninvolved in the matter appears between 0:14 and 0:22. Box their face for that window; record the parameters.
Exhibit: cctv_clip_03.mp4, 1280x720 Redact rectangle: X=410 Y=160 Width=140 Height=160 Start (s): 14 End (s): 22 Blur strength: 40 -> radius = min(40, floor(140/2)) = 40 Disclosure note: region (140x160 @ 410,160), 14.000s-22.000s, boxblur r=40, output H.264 MP4 CRF20, processed on-device.
Redact a document held up to the camera
A confidential document is visible from 1:02 to 1:09. Box the page generously and record the values.
Exhibit: interview_room.mov, 1920x1080 Redact rectangle: X=700 Y=300 Width=520 Height=680 Start (s): 62 End (s): 69 Blur strength: 50 -> radius = min(50, floor(520/2)) = 50 Output: interview_room.mp4 (audio copied, page illegible).
Two protected features in one exhibit (two passes)
A bystander's face AND an on-screen ID number both need redacting. One region per pass; document each separately.
Pass 1: exhibit.mp4 -> face box (window 14s-22s) -> step1.mp4 Pass 2: step1.mp4 -> ID-number box (window 30s-35s) -> final.mp4 Log both parameter sets; note exhibit was processed on-device.
Document the radius cap on a small region
When the protected feature is small, the radius is capped. Record that fact and enlarge the box if legibility remains, to keep the redaction defensible.
Region 80x40, strength 50 -> radius = floor(40/2) = 20 (capped) If still legible: enlarge to 160x100 -> radius = floor(100/2)=50 Note in disclosure: box enlarged to ensure full illegibility.
Mute a protected name spoken in the audio
An uninvolved party's name is said aloud. Blur the visual here, then mute that exact span with audio-mute-region, and record both steps.
Step 1: exhibit.mp4 -> redact face (14s-22s) -> visual.mp4 Step 2: visual.mp4 -> audio-mute-region (16s-18s) -> final.mp4 Disclosure: visual region blurred + audio 16.000-18.000 muted.
Edge cases and what actually happens
Protected feature moves out of the static box
By designThe box doesn't track motion. If a bystander walks across frame, one static box can't follow them. Either size the box to cover their whole path, or split into several time-windowed passes with a box per segment, and document each. For automatic face tracking, face-blur is the relevant tool — but note its detection is probabilistic, whereas a manual box is exactly recordable.
Small region remains legible after blur
Insufficient radiusThe boxblur radius is capped at half the smaller side, so a tight box may not fully obscure text. Enlarge the box (raising the cap) and document the enlargement, then re-verify on the frame where the feature is largest before disclosure.
Metadata not stripped from the exhibit
PreservedRedaction is a pixel operation; container metadata (timestamps, device, GPS) is untouched. If disclosure requires metadata removal, run the redacted file through the metadata-scrubber and document that step.
Protected audio left in the soundtrack
PreservedAudio is stream-copied. If a protected name or detail is spoken, blurring the video won't remove it — mute the span with audio-mute-region and record it.
Re-encode alters the exhibit's bytes
ExpectedThe video stream is re-encoded at CRF 20, so the output is not a stream copy of the original. This is inherent to applying a video filter. Keep the original exhibit sealed and unaltered; the redacted MP4 is a derivative produced by the documented process.
Output container differs from the source
By designA MOV or TS exhibit is output as MP4. If the disclosure format must match the original container, transcode the redacted result with the video-transcoder and note the conversion.
Box drawn past the frame edge
Crop errorIf X+Width exceeds frame width or Y+Height exceeds frame height, FFmpeg's crop fails. Keep the box fully within the frame; the picker clamps coordinates to the source dimensions to help avoid this.
Large multi-camera exhibit exceeds Free tier
Tier limitFree allows 1 GB per file. Long, high-bitrate exhibits may need Pro (10 GB) or Pro + Media / Developer (100 GB). Alternatively, redact each relevant segment after trimming with the lossless-trimmer.
Frequently asked questions
Is the redaction defensible and reproducible?
Yes. Every parameter — X, Y, Width, Height, Start, End, strength — is an explicit value (shown on the picker and in the fields) that maps directly to the FFmpeg crop/overlay/enable/boxblur arguments. Record them in your disclosure notes and the redaction can be reproduced exactly. The output codec is fixed (H.264, CRF 20).
Does the evidence ever leave my machine?
No. The source is loaded into an in-browser FFmpeg.wasm filesystem and processed on your device. No third-party server handles it, which supports chain-of-custody requirements.
Can the redacted feature be recovered?
No. The boxblur destroys the pixels in the region — it is not an overlay sitting on top. There is no original layer underneath the blur to reveal.
Does it modify the original exhibit?
No. A new MP4 is produced; the original file is left intact. Best practice is to keep the original sealed and treat the redacted MP4 as a documented derivative.
Why is the output not a stream copy of the original?
Applying a video filter requires re-encoding, so the video stream is re-encoded with H.264 at CRF 20 (audio is copied). This is inherent to any blur; document the codec parameters in your notes.
How do I redact a feature that moves across frame?
The box is static for its window. Either size it to cover the whole path, or split into multiple time-windowed passes and document each. Automatic face tracking via face-blur exists but is probabilistic; a manual box is exactly recordable, which often matters more for disclosure.
What about metadata in the exhibit?
Redaction doesn't touch metadata. If disclosure requires removing timestamps, GPS, or device info, run the redacted file through the metadata-scrubber and document that step.
A protected name is spoken aloud — how do I handle it?
Audio is stream-copied, so blurring the video won't remove spoken content. Mute the exact span with audio-mute-region and record the muted timestamps.
Can I redact more than one feature?
One rectangle per pass. Redact the first feature, then run the resulting MP4 back through the tool for the next, documenting each parameter set.
What input formats are accepted?
MP4, MOV, MKV, AVI and TS are recognised. The output is always H.264 MP4; transcode with the video-transcoder if the disclosure format must match the source container.
How large an exhibit can I redact?
Free allows 1 GB per file, Pro 10 GB, Pro + Media / Developer 100 GB, with no duration cap. For very large multi-camera exhibits, trim relevant segments first with the lossless-trimmer.
How do I document the blur strength on a small region?
Record both the strength you set and the effective radius min(strength, min(width,height)/2). If a small region was capped and remained legible, note that you enlarged the box to ensure full illegibility before disclosure.
Privacy first
Every JAD Video tool runs entirely in your browser via WebCodecs and FFmpeg (WebAssembly). Your video files never leave your device — verified by zero outbound network requests during processing.