I Built an AI That Edits My Videos Better Than I Do#

What if your raw footage could turn itself into a polished video while you sleep?

I used to spend 3-4 hours per video in a timeline editor. Dragging clips. Sliding audio tracks. Manually aligning cuts to the beat. For one video, fine. For ten videos a week, it's a death spiral.

Last month I ripped out that entire workflow. Now I edit videos by typing sentences.

The problem with timeline editors#

Traditional video editing hasn't changed much in 20 years. You open Premiere or DaVinci, drop footage onto a timeline, and manually click through everything.

This works for complex creative work. Color grading a short film? Yeah, you want fine-grained control.

But most of what I do isn't creative. It's repetitive:

• Trimming dead air from talking-head clips
• Adding subtitles to every video
• Cropping horizontal footage to vertical for TikTok
• Slapping background music on and ducking the audio when someone speaks

Each of these tasks takes 15-30 minutes in a GUI. Across a batch of videos, that's my entire afternoon gone.

What I built instead#

I turned video editing into a conversation.

I drop a file into chat. I describe what I want changed. I get the edited video back. No timeline, no menus, no mouse.

Here's what it handles:

• Trim, cut, and merge clips by describing timestamps in plain language
• Add background music with automatic audio ducking
• Generate and burn subtitles from speech (50+ languages supported)
• Color grade footage using natural descriptions
• Crop to vertical for TikTok, Reels, or Shorts
• Batch process multiple files with the same edit

The last one is what sold me. Batch processing 10 clips used to mean 10 separate editing sessions. Now it's one message.

The setup#

This runs on two skills from ClawHub: video-editor-ai for the core editing, and ai-subtitle-generator for caption work.

Install them:

clawhub install video-editor-ai
clawhub install ai-subtitle-generator

That's it. No API keys to juggle, no cloud services to configure, no Docker containers to babysit.

How it actually works in practice#

Single video edit. I drop in a talking-head clip and type:

Trim this video from 0:15 to 1:30, add background music (something upbeat),
and burn subtitles in English.

The agent parses that instruction, makes the API calls, polls for completion, and drops the finished MP4 back in chat. I don't touch a single button.

Batch edit. This is where the time savings get real:

I have 5 clips in /videos/raw/. For each one:
- Crop to 9:16 vertical
- Add auto-generated captions at the bottom
- Export as mp4

Five videos, one message. The agent processes them sequentially and delivers each file as it finishes.

I've run this on batches of 20 clips for a client's social media pipeline. What used to take me a full day now runs while I'm making coffee.

Subtitle workflow specifically#

Subtitles used to be my least favorite part of editing. Generate the SRT, import it, style it, position it, check for timing errors, repeat.

Now I drop a clip and say:

Burn subtitles in Spanish at the bottom, white text with black outline.

The ai-subtitle-generator skill handles transcription, timing, and burning. If I need the SRT file separately for YouTube's caption system, I just ask for it by name.

It handles 50+ languages. I've used it for English, Spanish, and Japanese content without issues.

Color grading without the technical jargon#

I'm not a colorist. I don't know what "lift the shadows 0.2 stops while pulling the highlights" means in practice.

With this setup, I describe the look I want:

Make it warmer, like golden hour sunset tones.

Or reference another clip:

Match the color grade of the first clip.

It works best with descriptive language rather than technical values. "Warm sunset tones" produces better results than "increase temperature by 15." The AI maps your description to actual grading parameters.

Things I learned the hard way#

Be specific about output format. "Export as mp4 at 1080p" gives you exactly that. Just saying "export" leaves the resolution ambiguous, and you might get a surprise file size.

Mention the source language if it's not English. The subtitle generator assumes English by default. For Spanish audio, say "burn subtitles from Spanish audio" to skip the wrong-language detection loop.

Don't upload sensitive footage unless you've reviewed the provider's data retention policy. This routes through external APIs. Faces, IDs, confidential screen recordings — think twice.

The real numbers#

Before this setup: ~3 hours per video for my standard edit (trim, subtitles, music, export). Across 10 videos a week, that's 30 hours.

After: ~15 minutes per video. Most of that is writing the instruction and reviewing the output. Across 10 videos, that's 2.5 hours.

I cut my editing time by roughly 90%. The remaining time goes to actual creative decisions — which takes to keep, what BGM fits the mood, whether the color grade matches my brand.

The repetitive work is gone.

What this isn't#

This isn't replacing creative editing. If you're cutting a narrative film, a music video, or anything where each frame matters, you still want a timeline and manual control.

This replaces the assembly line. The repetitive tasks that eat your time but don't require your brain.

If you're a YouTuber publishing 3+ videos per week, or a social media manager processing dozens of clips daily, that assembly line is probably most of your job.

Get started#

I published both skills on ClawHub so you can replicate this exact workflow:

• video-editor-ai — chat-based editing with BGM, subtitles, and export
• ai-subtitle-generator — auto captions, subtitle burning, SRT export

Install them, drop a clip, and type your first edit. It takes about two minutes to go from zero to your first AI-edited video.

More workflows like this are live at papayaclaw.com. I document the systems I actually use, not the ones I wish worked.