# Blur region (`blur_region`)
> Blurs one or more regions to hide things like emails, faces, or sensitive on-screen details. Regions are given in canvas pixels, with an optional time window.
- Area: [Studio editing](https://mcp.dadan.io/tools/studio-editing.md)
- Access: Makes changes; has an irreversible path, so clients ask to confirm (the explicit overwrite_original:true path irreversibly replaces the original's render)
- AI credits: yes
- Available on read-only connections: no

## Example prompt
"Blur the top-right corner from 0:10 to 0:30, an email shows there"

## Parameters
- `video_id` (string, required): The video's guid.
- `left` (number): Region origin X in canvas pixels.
- `top` (number): Region origin Y in canvas pixels.
- `width` (number): Region width in canvas pixels (required for a single region).
- `height` (number): Region height in canvas pixels (required for a single region).
- `shape` (string): Region shape: 'rect' (default) or 'circle'.
- `intensity` (number): Blur strength, 1–50.
- `start_seconds` (number): When the blur appears, in seconds. Omit for the whole video.
- `end_seconds` (number): When the blur disappears, in seconds. Omit for the whole video.
- `regions` (object[]): Several regions: {left, top, width, height, shape?, intensity?} each. Overrides the single region.
- `overwrite_original` (boolean): true = OVERWRITE the original video's render in place (irreversible). Default false = renders a NEW copy, original untouched.

## Description for AI assistants
Blur one or more regions of a video — hide emails, faces, or sensitive screen areas. Regions are in CANVAS PIXELS of the original recording: a single region via left/top/width/height (+ shape rect|circle, intensity 1–50), or several via regions. You cannot see the video, so NEVER guess where sensitive content is: ask the user for the region (get_video's width/height gives the frame size when available; it may be null — then ask the user for their recording resolution too). For privacy, a wrong-place blur is worse than none — when the user can't give a region, offer trim_video to cut the moment instead. start_seconds/end_seconds (SECONDS) limit when the blur shows — omit for the whole video. overwrite_original default false = renders a NEW copy, original untouched; true OVERWRITES the original's render. Returns a processing job — poll get_job_status; rendering takes a while.


## Known limitations

- Coordinates are caller-supplied. The model cannot see video frames, and get_video returns null width/height — so ask the user for the region or anchor to a transcript moment. A wrong-place blur silently ships the sensitive detail.
Canonical page: https://mcp.dadan.io/tools/blur-region
