Openclaw Skills reframe videos to new aspect ratios by extending the scene instead of cropping the subject.
Install & Download
1. ClawHub CLI
The fastest way to install a skill directly from the registry.
npx clawhub@latest install reframe-video
2. Manual Installation
Copy the skill folder to one of these locations
~/.openclaw/skills/ <project>/skills/ Priority: Workspace > Local > Bundled
3. Prompt Installation
Copy this prompt to OpenClaw to install it automatically.
Help me install reframe-video using Clawhub. If Clawhub is not installed, install it first (npm i -g clawhub).
Prefer to download?
Get the raw skill files in a ZIP archive.
What is Reframe Video?
Openclaw Skills' Reframe Video skill takes a finished clip, places it on a new canvas, and uses AI outpainting to generate the missing edges. That means a landscape master can become a vertical Reels or Shorts cut, or a square feed post, while the subject stays fully in frame.
This is designed for composition-preserving video reframing, not a simple crop and not a look-change edit. It relies on a live Ray 3.2 video model, supports asynchronous job execution, and is ideal when you need the target aspect ratio without sacrificing motion continuity or scene context.
Reframe Video Use Cases
- Convert a 16:9 landscape master into a 9:16 Reels, Shorts, or TikTok version without cropping the subject.
- Turn a finished clip into a 1:1 square post for feeds and galleries.
- Add sky, street, water, room, or other surroundings around a subject when the framing is too tight.
- Bias the composition with
settings.sourcePositionso the model opens space on the left, right, top, or bottom. - Extend an already polished clip after using
edit-video, then deliver the final social aspect ratio. - Create frame-extension variants even on the same aspect ratio by shrinking the source inside a larger canvas.
How Reframe Video Works
- Inspect the live schema for
luma:[email protected]withrunware-runorrunware-modelsso you can confirm the supportedwidth/heightpairs and the exactsettings.sourcePositionshape. - Provide the source clip as
inputs.video, either by upload or by URL. - Set
widthandheightto a supported pair that matches the target aspect ratio, such as a vertical or square output. - Write a
positivePromptthat clearly describes what should continue into the new area, so the model extends the scene instead of inventing unrelated content. - If needed, adjust
settings.sourcePositionto place or shrink the source inside the new canvas and bias where the extension appears. - Run
videoInferenceasynchronously, capture the returnedtaskUUID, and pollgetResponseuntil the job is complete. - Review the final
videoURL, then iterate with a clearer prompt or a different source position if the edge extension drifts.
Reframe Video Setup
No local installation is required; Openclaw Skills uses the hosted Runware workflow. Configure reframing like this:
- Make sure your workspace has access to
runware-runandrunware-models. - Confirm
luma:[email protected]is live and inspect the current schema before you submit a job. - Prepare a source video URL or upload, then choose a target
width/heightpair for the destination aspect ratio. - Write a
positivePromptthat explicitly names the sky, street, water, room, or other area that should be extended. - Optionally set
settings.sourcePositionwith fractionalx,y,width, andheightvalues to bias where the new space opens.
# Openclaw Skills / Runware preflight
# - access to runware-run and runware-models
# - source video ready
# - target canvas dimensions chosen
# - extension prompt prepared
#
# Reframing is asynchronous:
# submit videoInference, capture taskUUID, and poll getResponse until terminal
Reframe Video Data Schema & Taxonomy
Inputs and runtime metadata
| Field | Type | Required | Purpose |
|---|---|---|---|
inputs.video |
URL or base64 | Yes | Source clip to reframe |
width / height |
integer pair | Yes | New canvas size; must be a supported Ray 3.2 output pair and match the target aspect ratio |
positivePrompt |
string | Yes | Describes what should continue into the generated edges |
settings.sourcePosition |
object {x, y, width, height} |
Optional | Fractional placement and sizing of the source inside the canvas; values below 1.0 create space around the source |
taskUUID |
string | Output | Async job identifier returned by videoInference |
videoURL |
string | Output | Final reframed video |
Metadata taxonomy
- Composition metadata: target aspect ratio, source position, canvas bias.
- Generation metadata: model
luma:[email protected], prompt, async task state. - Delivery metadata: output URL, resolution tier, standard dynamic range, per-second billing.
Files and artifacts
- No local files are created by the skill itself.
- The primary deliverable is the final
videoURL. - The job is tracked through
taskUUIDuntil completion. - Supported output tiers run from 360p through 1080p, depending on the live schema and selected canvas pair.
Reframe Video Advanced Features
- Live schema inspection keeps Openclaw Skills aligned with the current
luma:[email protected]contract and avoids stale model assumptions. settings.sourcePositionlets you move or shrink the source to deliberately open space on the top, bottom, left, or right.- True outpaint-style edge generation preserves the subject instead of trimming it away.
- Same-aspect frame extension is possible by shrinking the source inside a larger canvas, not just by converting ratios.
- Asynchronous
videoInferenceplusgetResponsepolling fits long-running video jobs cleanly. - Pairs well with
edit-videoto lock the look first, then reframe as the final step. - Accepts either uploaded clips or source video URLs for flexible ingestion.
SKILL.md
Loading
Related Openclaw Skills
Runware Model Selector
Openclaw Skills helps agents pick the right live Runware model by matching the task outcome to verified capabilities, status, and price.
Product Ad Video
Openclaw Skills turns a product still into a polished promo video with cinematic motion, async delivery, and brand-safe control.
Talking Avatar
Openclaw Skills Talking Avatar turns a portrait, avatar, or existing video into a synchronized on-camera speaker from script or audio.
Restore and Upscale Images
Openclaw Skills restores blurry, noisy, or low-resolution images with the right model for sharper, more faithful results.
Photoreal Stills
An advanced AI generation skill designed to defeat synthetic rendering artifacts by prompting for optical behaviors, real-world camera properties, and intentional imperfections.
Logos and Vectors
Openclaw Skills creates real SVG logos or converts raster images into editable vector art that stays crisp at any size.
Featured*
