Let me describe my Tuesday evenings three months ago.
I'd shoot four to five short clips throughout the week — product demos, quick tutorials, a behind-the-scenes moment. Good content. Then Tuesday would arrive and I'd spend three to four hours staring at a timeline in DaVinci Resolve, doing the same things I'd done the Tuesday before: trim the dead air, add subtitles, resize to 9:16, drop in background music at a sane volume, export three versions for TikTok, Reels, and Shorts.
I wasn't editing. I was operating a conveyor belt.
The actual creative decisions took maybe twenty minutes. The rest was clicking through dialogs and watching export progress bars. I kept thinking: this should not require a human.
What I Found
I've been an OpenClaw user for about a year — I use it to manage my calendar, draft emails, handle some light scripting for my projects. It's basically a personal AI that runs locally and you can extend with community-built skills.
A few months ago I noticed a video editing skill on ClawHub. The description was straightforward: edit videos by chatting with your AI assistant.
I was skeptical. I've used enough AI video tools to know they usually mean "we'll describe what edits you should make" not "we'll actually make the edits." But I had a boring Tuesday coming up, so I tried it.
The short version: it actually makes the edits. You install the skill, drop a video file into your OpenClaw workspace, and start talking.
My Workflow, Step by Step
I've settled into a pretty consistent routine. Here's exactly how a typical video goes from raw footage to three exported files.
Step 1: Drop the file
I record on my phone or mirrorless, AirDrop to my laptop, and drag the file into ~/.openclaw/workspace/inbox/. That's the only manual step that stays manual.
Step 2: Open OpenClaw and start a session
I type one message to kick things off:
"New video in inbox:
demo-march-14.mp4. It's a 4-minute product walkthrough. I need it trimmed to under 90 seconds, subtitled, and exported in three sizes. Let's go."
OpenClaw picks up the file, calls the video editing skill, and we're off.
Step 3: Let it run
From here it's a back-and-forth, but it goes fast. The skill knows how to handle the heavy lifting — it sends the video to the processing pipeline, streams back status updates, and when it's done it tells me exactly where the output landed.
The full session usually takes 8–12 minutes per video depending on length. I keep my laptop open but I'm not watching it. I'm making coffee.
The Prompts I Actually Use
This is the part I wish someone had given me when I started. Here are my most-used prompts, copy-pasteable:
For trimming + compression:
Trim the first 8 seconds and the last 5 seconds from [filename]. Then compress to under 50MB without dropping below 1080p.
For subtitles:
Add auto-generated subtitles to [filename]. Burn them in at the bottom third, white text with a subtle black shadow, no background box.
For multi-platform export:
Take [filename] and export three versions: one at original aspect ratio for YouTube, one cropped to 9:16 at 1080x1920 for TikTok and Reels, one at 1080x1920 with max 60 seconds for Shorts. Name them [basename]-youtube, [basename]-vertical, [basename]-shorts.
For background music:
Add [music-file.mp3] as background music to [filename]. Set the music volume to 15% and duck it to 5% whenever speech is detected.
For a full batch:
Process all .mp4 files in the inbox folder: trim 5 seconds from the start of each, add auto-subtitles, export as 9:16 vertical. Move originals to /inbox/processed when done.
That last one is my favorite. I drop five clips, send one message, go to sleep. In the morning there are fifteen files (three versions each) waiting for me.
The Numbers: Before vs After
I tracked my time for six weeks before switching and six weeks after. This isn't marketing copy — this is what I logged.
| Task | Before (per video) | After (per video) |
|---|---|---|
| Trimming + cleanup | 12 min | ~0 (automated) |
| Subtitle generation + review | 18 min | 4 min (review only) |
| Resize + format variants | 15 min | ~0 (automated) |
| Music + audio balance | 10 min | ~0 (automated) |
| Export management | 8 min | ~0 (automated) |
| Total per video | ~63 min | ~8 min |
With four videos a week, that's roughly 220 minutes back every week. About 3.5 hours.
Cost-wise: the skill runs on credits. I'm paying roughly $12–15 a month at my current volume. Before I was paying in time, which at any reasonable rate is worth way more than fifteen dollars.
Honest Limitations
I'd be lying if I said this was perfect. Here's what still trips me up:
Subtitle accuracy on technical jargon. If I'm talking about specific software terms or product names, the auto-generated subtitles will sometimes mangle them. I still do a quick manual review on anything that's going to be public-facing. Takes about 3–4 minutes per video.
Music ducking isn't always perfect. The speech detection for auto-ducking works well on clean recordings. If I shot in a noisy environment or there's significant background noise in my source, it can duck at weird moments. I've learned to specify stricter ducking parameters when this is a risk: duck to 3% when speech confidence is above 80%.
Large files and timeouts. Anything over about 800MB can sometimes timeout during upload depending on my connection. I've started transcoding very long source files first with a quick local ffmpeg pass to h264 before sending them. Not ideal, but it's a one-liner.
The skill can't watch your screen. This is OpenClaw operating on files, not a GUI agent. If your workflow depends on real-time visual feedback during editing (color grading, precise subtitle positioning, etc.), you'll still need to do that part yourself.
How to Get Started
If you're already an OpenClaw user, setup is about five minutes:
- Go to ClawHub and search for a video editing skill
- Install the skill:
npx clawhub@latest install video-editor-ai --force
- You'll be prompted to register for an API key when you first use it (there are free credits to test with)
- Drop a video in your workspace and start talking
Source code is on GitHub if you want to see how the skill is structured.
If you're not yet an OpenClaw user — it's a local AI assistant you run yourself, kind of like having a personal Claude or GPT that also has access to your filesystem, your apps, and community-built skills. You can find it at openclaw.ai.
Final Thought
I still enjoy video editing when it's actually editing — when I'm making a real creative decision about pacing or structure or what to cut. What I don't enjoy is the mechanical repetition that comes after.
OpenClaw with a good video editing skill drew a clear line between those two things. The creative stuff stays with me. The repetitive stuff goes to the machine.
Three and a half hours a week adds up. I've spent that time shooting better content instead.
If you've got questions about specific prompts or use cases, drop them in the comments — happy to share what's worked.
Related skills from the same backend:
-
Shorts Editor — vertical video for TikTok, Reels & Shorts. Install:
npx clawhub@latest install shorts-editor -
Video Caption Tool — auto subtitles, 50+ language translation, SRT export. Install:
npx clawhub@latest install video-caption-tool
Top comments (0)