AI & Workflow · 6 min read
Working with AI in the edit suite: my real workflow
Published May 2026
People keep asking me if AI is going to take editing jobs. The truer question is whether AI is already changing editing jobs — and the answer there is yes, in ways I didn't expect when I first plugged Runway into my pipeline a year ago.
I cut around 40 short‑form videos a week with an eight‑hour turnaround per piece. That math only works if every step of my workflow earns its place. Here's where AI actually does that, and where it still doesn't.
Where AI saves me real time
Transcription & rough cut. Descript turns a one‑hour interview into a navigable text document in three minutes. I cut 40% of the footage by deleting words on a page before I open Premiere. The cut I bring into the timeline is already 70% of the way to a first pass.
VO scratch. ElevenLabs gives me a usable voice‑over in 30 seconds. I'd never publish a synthetic VO for a client without permission, but for client review cuts and rough timing, it's a different world.
VFX previs. Runway lets me show a concept in five minutes that used to need a half‑day in After Effects. We move from "could we do this?" to "do we like it?" much faster, which is where the real creative work happens anyway.
Where I still want a human in the room
Hooks. Music selection. Pacing decisions. The emotional curve. Anything where the answer to "is this right?" lives in the gut, not the spec sheet. AI helps me get there faster, but the call is still mine, every time.
The honest tradeoff
AI compresses the boring parts of editing. That's a gift. But it also raises the floor — which means the bar for what counts as "good editing" is moving up. The advantage now isn't speed. It's taste. Build the taste; the tools are the easy part.
AI & Workflow · 5 min read
ElevenLabs made my voice work twice as fast — here's how
Published April 2026
Three things used to eat my Friday afternoons: re‑recording client VO, fixing room tone, and waiting for the talent to be free again. ElevenLabs didn't kill those problems for me. It moved them.
What I actually do with it
Scratch tracks. Every cut starts with a placeholder VO. With ElevenLabs I can generate a final‑quality scratch in seconds, locked to my edit's timing. When the real talent records later, I'm not editing to a guess.
Pickups. A client sends a script revision at 11pm. I drop the new line into the timeline, the cut still works, the client signs off the next morning. We go to actual VO once on the version that's actually approved.
Multilingual reels. A travel short I cut for a Chinese audience needed a Mandarin reading. I generated a clean reference, sent it to a native speaker for review, and shipped a polished version a day faster than I would have otherwise.
The quality bar
Synthetic voice isn't magic — it can sound flat if you don't push it. I always: write for the model (shorter sentences, clear punctuation), generate three takes per line, and pick the one with the right breath. If a client wants a recognisable human voice attached, we use a real one. AI is the scratch and the pickup, not the identity of the brand.
One ethical line
I don't clone real people's voices without their explicit, written permission, and I tell every client when synthetic voice is in the cut. The tool is fast. Trust isn't.
Storytelling · 4 min read
Why the first two seconds decide everything
Published March 2026
I've watched analytics for thousands of short‑form videos and I keep seeing the same shape: a cliff at the two‑second mark. If a viewer is still there at two seconds, they usually finish. If they leave at one and a half, they were already gone before your hook had a chance to speak.
The hook isn't a tagline. It's a contract.
You're saying to the viewer: stay here for the next 30 seconds and I'll give you something specific in return. If you don't name what that thing is in the opening frame, you've broken the contract before you signed it.
What I do in the first two seconds
One visual idea, no setup. The first frame should already be the most interesting thing on screen. No B‑roll establishing shot. No logo. No "hi, today I want to talk about." Open on the answer to a question the viewer didn't know they were asking.
One sound that doesn't sound like everyone else. The platform feed is a wall of similar audio. A small percussive hit, a vocal whisper, a pause where there should be music — anything to break the rhythm of the scroll.
One promise, in one line. If your hook needs two sentences, it isn't a hook yet. Cut until it fits.
The retention curve doesn't lie
When the curve is shaped like a cliff at two seconds, the hook is wrong. When it's shaped like a slow slide, the pacing is wrong. When it's shaped like a smile — flat through the middle and a small bump at the end — you've nailed both. That's the cut you publish.
Viral Mechanics · 7 min read
Hook, cut, climax: a three‑beat recipe for retention
Published February 2026
Most viral short‑form looks chaotic on the surface and follows a strict structure underneath. After cutting hundreds of pieces for million‑subscriber channels and brand accounts, I keep coming back to the same three‑beat pattern: hook · cut · climax. Here's what each beat is actually doing.
Beat one: the hook (0–2 seconds)
This is the contract. State the thing the viewer will get if they stay. Use specifics, not categories — "I cut 40 videos a week" beats "I'm a busy editor." Visual and audio land in the same frame. No setup.
Beat two: the cut (2–18 seconds)
This is the body. Most people over‑edit it. The trick is not "more cuts" — it's useful cuts. Each one should either move the story forward, change the emotional register, or reveal new information. If a cut isn't doing one of those three things, delete it.
Pacing matters more than density. A held shot with the right music breathes. A held shot with no purpose dies. Watch your cuts at half‑speed once a week and you'll see which ones earn their place.
Beat three: the climax (last 5–8 seconds)
The climax is the payoff to the contract you signed in the hook. Not a CTA. Not a logo. The thing the viewer was waiting for. Land it cleanly, give it one extra beat to sit, then end. The end is your most underrated edit decision.
Why this works
Because it mirrors how attention actually behaves. Curiosity opens a loop, motion holds it, payoff closes it. When you respect those three beats, retention goes up across every platform. When you skip any one of them, the platform notices before your viewer does.
The exception
Comedy. Comedy lives or dies on rhythm, and rhythm is its own structure. But for branded content, social proof, education, and most cause work — hook, cut, climax holds up week after week. Build the muscle. Ship the work.