Introduction

If you are processing video materials in DSH or agent development workflows, you typically need to first cut the video into shots or keyframes, then let the model output editable text based on the visuals, and finally export it into documents or tables. dsh-bcc puts this kind of work into DeepSeek Harness’s web session: it provides a “Package Breakdown” panel, a set of video decomposition tools, and the ability to export results.

Below is an introduction to its positioning, installation methods, and two main pipelines.

Positioning

dsh-bcc is maintained by aixlb with an MIT license.

Its one-sentence positioning is:

Package Breakdown for DeepSeek Harness: video → script / storyboard / style guide.

It handles ffmpeg cut detection, keyframes, and exporting to docx, xlsx, or html. Analysis uses the current DSH session model; there is no Gemini/Kimi configuration page, nor does it bundle an API key.

Core Capabilities

dsh-bcc is mainly aimed at “video input, text output” workflows:

  • Convert video into script, storyboard, or style guide within DeepSeek Harness.
  • Handle ffmpeg cut detection, keyframes, and export to docx, xlsx, html.
  • Analysis uses the current DSH session model without bundling an API key.
  • The recommended model is deepseek-v4-flash-vision-exp; if the current model cannot view images, the plugin will prompt you to switch in the DSH model picker and will not rewrite the default model.
  • Supports Script Breakdown interval slicing and uses extract/merge prompts.
  • Supports Shot Breakdown smart/scene cuts and combines with shot style from the panel.
  • Provides command palette entries: /bcc-script, /bcc-storyboard.

The tools provided by the plugin include:

bcc_probe
bcc_shots
bcc_sample_script
bcc_set_cuts
bcc_read_frames
bcc_script_coverage
bcc_export
bcc_master
bcc_project_list
bcc_project_open

Installation and Activation

Runtime requirements:

node ^22.19.0 || >=24.0.0

Installation command:

npx @deepseek-ai/dsh plugin --profile web add github:aixlb/dsh-bcc

After installation, restart dsh web and hard refresh the page in your browser.

After entering a session, click “Package Breakdown” at the top of the session header to expand the project panel.

Typical Usage

The basic steps are as follows:

  1. In the “Package Breakdown” project panel, select or upload a video.

  2. Select Script Breakdown or Shot Breakdown.

  3. Follow the corresponding pipeline according to your goal.

Script Breakdown Pipeline

Script Breakdown uses interval slicing and extract/merge prompts. The pipeline order is:

bcc_shots cutMethod=interval → bcc_read_frames source=shots → merge beats → bcc_script_coverage → bcc_export

Here, bcc_read_frames source=shots needs to run until remaining=0 before proceeding with subsequent merging and exporting.

Shot Breakdown Pipeline

Shot Breakdown uses smart/scene cuts and combines with shot style. The pipeline order is:

bcc_shots → bcc_read_frames source=shots → bcc_export

Command Palette

Can also be used in the command palette:

/bcc-script
/bcc-storyboard

Applicable Scenarios and Notes

Suitable for scenarios where you need to process video materials in a DSH web session and want to break down the video into scripts, storyboards, or style guides.

Pay attention to the following when using:

  • Analysis depends on the current DSH session model; if the current model cannot view images, the plugin will prompt you to switch but will not rewrite the default model.
  • The recommended model is deepseek-v4-flash-vision-exp.
  • Script output focuses on visuals; dialogue comes from burned-in captions; if dialogue is not seen, it will be written as 【画面未见对白】.
  • The plugin runs with the permissions of the current DSH process; you should check the source code and license before installing.
  • The license is MIT.
  • It is a community plugin within the DSH plugin ecosystem; the plugin directory is an independent site and does not form an official affiliation with DeepSeek / Firmirrors.

Links

  • GitHub: https://github.com/aixlb/dsh-bcc
  • Directory page link: https://www.skillhub.cn/plugins/aixlb/dsh-bcc (This address comes from the plugin link and has not been verified in the materials scraped for this article)