Introduction

Splitting a long video into multiple segments based on scene changes is a common requirement in video editing, asset archiving, and batch analysis. Manually performing this task usually involves two approaches: first, dragging the progress bar frame-by-frame to find cut points and then manually splitting with ffmpeg, which is time-consuming and prone to missing shots; second, writing a scene detection script, which requires handling details like detection algorithms, first/last frame export, and segment duration alignment yourself, often leading to the pitfall of “duplicate frames between the end of the previous segment and the start of the next.”

The bozo-scene Skill encapsulates this process into a single step. The following section introduces what it is, what it can do, and how to enable it.

What is this?

bozo-scene is a local video storyboard (scene) analysis and splitting tool, identified as @indiv-bozoyan/bozo-scene. It is maintained by the community publisher user_6419d865, categorized under design-media, and is currently at version 1.0.0.

It utilizes PySceneDetect to automatically detect scene transition points in videos, automating the step of “finding cut points,” and provides two types of splitting capabilities based on this. The entire tool runs locally; there is no need to upload videos or configure an API key (tag: requires_api_key=false).

Core Functions

It addresses two types of tasks, both centered around storyboarding/scenes.

  1. Storyboard First/Last Frame Extraction: After detecting each scene, it exports a first-frame image and a last-frame image for every scene. This is suitable for scenarios requiring quick previews of video scene structures, generating storyboard thumbnails, or creating asset indexes.

  2. Storyboard Video Export: It cuts each scene into an independent MP4 segment. A notable detail here is that it automatically subtracts one frame’s duration during splitting to avoid the issue of “the last frame of the previous segment equaling the first frame of the next segment” (overlap between adjacent segments). Without this step, the split segments would have a duplicated frame at the boundary, causing a visual flicker during playback if concatenated.

Triggering: Officially, this Skill should be used when the user’s intent involves: mentioning “storyboard,” “scene switching,” “shot switching,” “video splitting,” “cut video by scene,” “extract scene first/last frames,” “split scene video,” “storyboard preview images,” “cut video by visual changes,” or “segment by shots,” or when the user needs to cut a long video into multiple segments based on scene changes or batch analyze video scene structures. Even if the user does not explicitly mention PySceneDetect, as long as the intent is to split a video by visual changes or analyze the number of scenes in a video, this Skill should be prioritized over asking the user to manually write ffmpeg commands.

Installation and Activation

bozo-scene uses the universal SKILL.md format and is distributed via SkillHub.

Official Address: https://api.skillhub.cn/indiv-bozoyan/bozo-scene
SkillHub Page: https://www.skillhub.cn/skills/indiv-bozoyan/bozo-scene

Note: The official materials retrieved this time did not include specific installation commands, directory paths, or configuration fields. Since the official page was inaccessible during retrieval, it is recommended to verify the accurate installation method for your specific editor (Cursor, Codex CLI, Claude Code, etc.) at the official address above before enabling. Do not copy third-party commands directly.

Based on its distribution form, it originates from the community, belongs to the design-media category, and is tagged as requiring no API key. Once enabled, it can perform storyboard analysis on local video files immediately.

Typical Usage

Since the official materials did not provide reproducible command-line examples, the following section only describes the operational intent corresponding to its capabilities. Specific commands should be based on the official page.

  • When the user expresses “cut this video by scene,” the Skill will first perform scene detection on the video, then export each scene as an independent MP4 based on the detected cut points.
  • When the user expresses “extract scene first/last frames” or needs “storyboard preview images,” the Skill will export one first-frame image and one last-frame image for each scene.

Both tasks are based on the same scene detection results; the only difference lies in whether the output is images or video clips.

Use Cases and Notes

Suitable Scenarios:
* Splitting long videos into multiple asset clips based on visual changes.
* Batch understanding the scene structure of one or more videos.
* Generating first/last frame previews for each scene to create asset indexes.

Points to Note:

  1. Detection Accuracy: It relies on PySceneDetect for scene detection. The accuracy of cut points depends on the intensity of visual changes in the video itself. For videos with smooth changes or soft transitions, detection results may not meet expectations, and manual review of cut points may be necessary.
  2. Frame Subtraction: The automatic subtraction of 1 frame during splitting is intended to prevent duplicate frames at the boundaries of adjacent segments. If your use case has special requirements for frame-level precise alignment (e.g., for subsequent frame-by-frame comparison), please check if this behavior meets your expectations.
  3. Performance: The tool runs locally. Processing large or long-duration videos depends on machine performance and PySceneDetect parameters. It is recommended to verify the effect with a short video first.
  4. Configuration: The retrieved official materials did not provide detailed parameter configuration instructions. If you need to adjust detection sensitivity or output format, you need to check the official page to confirm support.

Conclusion

The value of bozo-scene lies in transforming the task of “splitting video by visual changes” from manual progress bar dragging and hand-writing ffmpeg commands into a single call. It automatically detects scene cut points, supports storyboard first/last frame extraction and storyboard video export, and handles the common issue of frame overlap between adjacent segments during splitting.

Official Address: https://api.skillhub.cn/indiv-bozoyan/bozo-scene