Script to video, using footage you actually filmed

Search for "script to video" and nearly every result works the same way: paste in some text, and the tool fills it with stock clips, AI-generated scenes and a synthetic voice. That's great for an explainer about something you never filmed. It's the wrong tool if your phone is already full of footage and you want a video that actually looks like you.

The short answer

Write the script as a few short lines, one idea each, and use an app that matches each line to your own clips. RawCut reads every clip on your iPhone once: what's on screen second by second, and what's said. When you write "the first coffee of the morning," it finds the seconds where that actually happens, not just a file with a similar name, and cuts them into a vertical 9:16 video in the order you wrote them.

Three kinds of script-to-video tools

Everything sold as "script to video" falls into one of three buckets, and the difference is where the footage comes from:

• Stock-based. Your script is matched against a licensed library. Fast and clean, but the same shots show up in thousands of other videos, and viewers can tell.
• Generative. Every shot is created by an AI model. Great for things that can't be filmed, but it's often labeled or watermarked, and it rarely feels personal.
• Your own footage. Your script is matched against clips you filmed. Nothing is invented, so the result is only as good as what you shot, but nobody else can make the same video.

If you're a creator, a small business or anyone with a camera roll full of real moments, the third kind is usually the one you want. It's also the hardest one to build, which is why there are so few of them.

Why matching your own footage is harder

Matching a sentence to a stock library is easy: the library is tagged, the clips are short and clean, and any shot of "a coffee cup" will do. Your footage is different. Your clips are long, untagged, and shaky in places, and the good two seconds are buried in the middle of a two-minute recording.

That's the problem RawCut is built around. It reads each clip in detail when you add it, not when you ask for a video: what's shown, what's said, the setting, the mood, how steady the camera is. By the time you write a script, the hard part is already done, and the same library answers every script you write after that.

How it works in RawCut

1. Add clips once. Pick them from Photos. Each one is read while the app is open and stays ready for every video you make after that.
2. Write the script. A single sentence, or a few lines. A one-line theme is expanded into shots for you.
3. Get the cut. Every line is matched to the moments that answer it, and the cut follows your lines in the order you wrote them.
4. Fix anything you don't like. Replace a shot, trim it, reorder, mute a clip, add on-screen text, or turn on automatic captions.
5. Export. A vertical video for TikTok, Reels or Shorts.

How to write a script your footage can answer

The script is your only steering wheel, so it's worth an extra minute. These habits make the difference between a cut that lands and one that guesses:

Describe what a camera could see. "Rain on the window" is answerable. "A feeling of calm" isn't, because no clip contains a feeling. Put the mood in on-screen text and keep the picture in the line.

One idea per line. "Walking to the farmers market and buying bread" is two shots pretending to be one. Split it, and each half gets its own best moment.

Use your own words. Speech is read too. If you said something good on camera, write it close to how you said it, and the cut can land on that sentence and keep it whole instead of chopping it mid-word.

Name what you actually filmed. If the clip is of a pier, write "pier," not "ocean view." The closer your words are to what is really on screen, the less the match has to stretch.

Let the order tell the story. The cut follows your lines from top to bottom. Before and after, start to finish, problem and fix: pick a shape and the video will have one too.

Keep it short. Four to eight lines suit a 15- or 30-second video, which is where most short-form holds attention anyway.

Three example scripts

A Saturday, from everyday footage:

1. Early light through the kitchen window
2. Coffee being poured
3. Walking to the farmers market
4. Picking out tomatoes
5. "This is the best part of my week"

The first four are visual and specific. The last one is a line said to camera, so the cut uses the moment it was spoken.

A small business, from a week of behind-the-scenes clips:

1. Unboxing a delivery of raw materials
2. Hands measuring and cutting
3. The finished product on the workbench
4. Packing an order
5. Taping the box shut

A process video like this is one of the easiest formats to cut from existing footage, because most people film these steps without thinking about it.

A faceless b-roll video, with no one talking to camera:

1. City street at dusk
2. Neon sign
3. Rain on a car windshield
4. Headlights passing

Every line here is a scene, so RawCut cuts it from footage without faces. See the guide to faceless videos from your own footage for more.

Script length and pacing

Pick a length (15, 30 or 60 seconds) and the cut aims for it; your lines decide what fills it. As a rough guide, a line that's answered by b-roll gets a few seconds, and a line you said on camera runs as long as the sentence takes to say. So:

• A 15-second video: three to five b-roll lines.
• A 30-second video: six to eight lines, or one spoken line with b-roll around it.
• A 60-second video: better built around something you said, since pure b-roll rarely holds a full minute.

If the cut feels long, delete a line rather than trimming every shot. One fewer beat almost always beats eight rushed ones.

When a line has no match

Sometimes nothing in your library answers a line. Generators never run into this, because they just invent the shot. RawCut won't: it leaves that line out of the cut and tells you which one it couldn't answer.

That turns out to be more useful than it sounds. A shot that almost fits is how a video stops making sense, and knowing exactly which shot is missing tells you what to film next time. You can also rewrite the line to match something you do have, or just drop it.

Adding your voice and on-screen text

Some ideas don't have a picture. That's what on-screen text is for: put the claim, the price or the punchline in text over a shot that supports it. If you'd rather say it, export and record a voice-over in TikTok or Instagram. RawCut never generates a voice for you.

If your script uses lines you said on camera, turn on automatic captions. The words are pulled from your clips and timed word by word, so viewers watching with the sound off still get the sentence.

When a generator is the right tool

To be fair: if you have no footage of the subject, or you're making an explainer about something you could never film, a text-to-video tool or a stock library is the right call, and RawCut can't help. It needs your clips. The trade-off is that everything in the finished video is real, and nobody else has the same shots.

Film now, script later

The best scripts come from a well-stocked camera roll. A few habits make your footage easier to write for later: hold each shot for five seconds or so, film the "boring" in-between shots (doors, hands, streets), and say the important thing out loud on camera at least once. You won't need a plan when you're filming; the script can come weeks later.

Related

• Make a video from your own clips
• Find a specific moment in hours of footage
• How to choose a camera-roll video app
• Make an Instagram Reel from your iPhone videos

Questions

Free to try. iPhone, iOS 15 or later.