How to Use ScreenKite AI Video Editing: Complete Workflow from Recording to B-Roll

In our previous post, we introduced ScreenKite's AI editing capabilities.

This guide goes deeper with a complete step-by-step walkthrough of the entire AI editing workflow, covering:

• Prerequisites and setup
• Automatically cutting spoken tracks and pauses
• Automatically generating contextual B-roll
• What actually happens under the hood at each step

If you want to know how this workflow operates in practice, here is everything you need.

Before You Start: What You Need

ScreenKite's agentic editing workflow leverages local-first developer tools. Make sure you have the following ready before starting:

1. Install ScreenKite

Download and install ScreenKite from screenkite.com.

2. Install an AI Coding Agent CLI

ScreenKite's editing commands are executed through your choice of coding agent CLI:

• Claude Code
• Gemini CLI
• OpenAI ChatGPT Codex

These tools run locally on your Mac, and ScreenKite exposes typed CLI commands and MCP tools for them to interact with your project timeline.

3. Configure Speech Transcription

Automated spoken cuts require word-level audio transcription. You have two options:

Option 1: ElevenLabs API

Create a .env file in your project directory with:

ELEVEN_LABS_API_KEY=your_api_key_here

Option 2: Local WhisperKit Model

In ScreenKite's Settings → Transcription, download a WhisperKit model. This runs 100% offline with zero API keys required.

4. Install B-Roll Generation Dependencies (Optional)

If you plan to generate Remotion-based animated B-roll overlays:

• Node.js (for Remotion rendering)
• uv (Python package manager)

Step 1: Record Your Screen Video

Launch ScreenKite and record your screencast.

Once finished, ScreenKite saves a .skbundle package. This bundle contains your screen capture, webcam stream, microphone, and internal system audio tracks.

Step 2: Open the Project with Your AI Agent

Open the ScreenKite editor and click the AI Agent button in the top left.

Your AI tool (such as Claude Code) connects directly to the project session.

The agent inspects the project structure:

• Video duration
• Available tracks (screen recording, webcam, microphone, system audio)
• Existing auto-zoom keyframes
• Overlay and caption track states

Step 3: Automatically Cut Spoken Audio

Tell the Agent What You Want

Simply instruct the agent in plain natural language:

"Remove silent pauses and filler words from the audio track."

What the Agent Does

1. Extract Audio & Transcribe: Extracts the microphone track and passes it to ElevenLabs or local WhisperKit for word-level alignment.
2. Detect Dead Air: Analyzes audio waveforms to identify pauses and silences.
3. Detect Filler Words: Scans the transcript for filler phrases ("um", "uh", "like").
4. Preview Cut Plan: Presents a clear list of proposed cuts for your approval before making any changes.

Review and Apply

Confirm by saying "cut" or "apply". The agent executes the trim operations directly on the timeline.

Step 4: Automatically Generate B-Roll Overlays

Tell the Agent What You Want

"Generate contextual B-roll elements based on the video transcript."

Content Beat Analysis

The agent segments your video into narrative beats and proposes visual components:

• Intro Beat: Channel nameplate, category tag
• Feature Overview: Highlight cards, constraint pills
• Core Demo: Feature chips, live timer badges
• Summary: Action checklist card

Choose B-Roll Density

• Sparse: 4 key visual elements (clean, documentary-style)
• Medium: 7 balanced elements (recommended default)
• Dense: 10 visual elements (ideal for dense technical tutorials)

Render and Place on Timeline

The agent renders the components as video assets and places them directly onto ScreenKite's Overlay track with smooth motion transitions.

Privacy & Local-First Processing

• Video source media remains strictly on your local disk.
• Coding agents run locally on your Mac.
• Transcripts and timeline manifests reside in your project bundle.
• If you use WhisperKit, transcription is completely offline.

Summary

ScreenKite's AI video editing is transparent and controllable. It lets you direct your edits using plain language while keeping full oversight over every trim, zoom, and overlay.

Download ScreenKite and experience agentic video editing on macOS today.