The Video Remix Agents are a suite of three AI-powered agents on the Pencil platform. Together, they form a sequential pipeline that transforms raw video into polished short-form content.
Prerequisites:
Access to your Pencil workspace.
Raw video files in MP4 format.
If you don't see them on the main Agents tab in Pencil, you can run a search in the search bar for "video rem" to locate them.
The Three-Stage Pipeline
We recommend running the agents in this specific order:
Video Splitter Agent > Video Script Agent > Video Stitch Agent.
Stage 1: Video Splitter
What it does: Intelligently splits raw video files into targeted segments and generates a scene manifest with exact timestamps and metadata.
Input: Raw MP4 videos.
Output: A scene manifest containing file names, scene IDs, and start/end timestamps.
Standalone Use Case: If you just need to rapidly extract the best scenes from a massive raw shoot without storyboarding, you can use this agent on its own to log and clip your footage.
Stage 2: Video Script
What it does: Reads the scene manifest to write a structured storyboard.
Using Templates: You can optionally select a video template to control the scene count, layer positioning, and duration.
Output: A formatted storyboard with scene breakdowns, text layer content, and a compliance checklist.
NB! When working with Templates in this way, the template must have only one video layer per scene.
Standalone Use Case: If you already have a manual list of video clips and timestamps, you can feed them directly into this agent to rapidly brainstorm narrative structures and generate multiple storyboard variations for A/B testing without committing to rendering them yet.
Stage 3: Video Stitch
What it does: Takes the storyboard and stitches the referenced clips into a final video.
Output: A cohesive, rendered video with all clips and text overlays properly placed.
Chat vs. Workflow Modes
You can utilise these agents in two distinct ways depending on your needs:
Standalone Use Case: If your creative team prefers to manually write their own scripts or storyboards using exact timestamps from existing assets, you can feed that manual script directly into the Stitch Agent to automatically assemble and render the final video, bypassing the AI scripting phase entirely.
Feature | Chat Mode | Workflow Mode |
Experience | Interactive, conversational | Visual, node-based canvas |
Best For | Exploration and testing | Automation and scaled pipelines |
Data Flow | Manual stage transitions | Automatic data flow between nodes |
Capacity | Max 14 videos per session | Supports larger collections |