You've shot the footage, you've got your script, and now you have to edit. That's usually where projects die: cutting the silences, syncing the subtitles, finding music, exporting in the right format for TikTok and then again for YouTube. AI now handles a good chunk of that work, but not all of it. It's very effective on repetitive tasks (transcription, subtitles, removing dead air, auto-reframing, audio cleanup) and noticeably weaker on pacing, intent and shot selection — which is exactly what makes a video enjoyable to watch.
This article walks you through three concrete ways to edit a video with AI, what each one costs, how long it really takes, and a six-step workflow you can reuse on every video you make. No technical skills needed: if you can send a file from your phone, you can start today.
What AI can do in a video edit (and what it gets wrong)
AI handles the mechanical parts of editing very well: transcription, subtitles, cutting silences, reframing, audio cleanup. It handles pacing, intent and shot selection badly — and those stay your job.
An edit isn't one single action, it's a series of micro-tasks. Here's which ones are reliable today and which still need your eye.
| Task | What AI does | Real-world reliability |
|---|---|---|
| Transcription and subtitles | Generates text synced to the voice | Good: expect a few fixes on proper nouns and technical terms |
| Cutting silences | Removes dead air, hesitations and “ums” | Good, but review it: it sometimes cuts breaths you actually wanted |
| Reframing 16:9 to 9:16 | Tracks the face or the main subject | Good with one person on screen, unpredictable with several |
| Audio cleanup | Removes background noise, echo, hiss | Excellent — this is the most immediate win |
| Generating shots (B-roll) | Creates images or cutaway sequences | Mixed: you have to sort through it and often regenerate |
| Shot selection and pacing | Offers a fully automatic edit | Weak: the result feels flat, the calls are yours to make |
The practical takeaway is simple. On a two-minute short video, AI absorbs the grunt work (transcribing, subtitling, cleaning, resizing) in a few minutes, whereas those steps easily eat an hour by hand. But if you also hand it the decision of what stays on screen, you end up with a video that's technically fine and boring to watch. The right instinct: AI for execution, you for the decisions.
A quick word on vocabulary, because it causes a lot of confusion. “AI video editing” actually covers two different things: edit assistance (you have footage, AI helps you assemble it) and video generation (you filmed nothing, AI creates the images from a text prompt). These are two distinct crafts, and often two distinct tools. If it's the second one you're after, start instead by understanding how to generate a video from text.
Method 1: edit in an environment where every AI tool is already in one place
The most efficient route for a beginner is to work in a space where the AI tools are already connected and paid for: you generate the shots, the voice-over and the music in the same place, then edit — without stacking up five subscriptions.
The real barrier when you're starting out isn't technical difficulty, it's accumulation. An image generator, a video generator, a voice-over tool, royalty-free music, a subtitle tool: each one bills between €15 and €30 a month, and you're at €100 a month before you've published a thing. That's on top of the time spent creating accounts, learning interfaces and moving files from one to the next.
That's the idea behind Skilzy's AI Lab: the tools are built into the platform with credits included. You follow a program that walks you through the method, and you apply it immediately in the same tab, on your own project. Concretely, for a short video aimed at Instagram or TikTok, you go: script, shot generation, voice-over in French, music, then assembly and subtitles.
The platform has more than fifteen programs (image creation, UGC videos, faceless YouTube channels, talking avatars, voice-over, music, automation with n8n, prompt writing, AI-assisted coding). For editing social videos, the most direct entry point is the program on creating UGC videos with AI, which takes you from the script all the way to the published video, vertical format included.
Two concrete things before you dive in:
- Try it without a credit card. The discovery demo gives you 7 days with 1 image, 1 video, 1 music track and 10 messages. You do have to create an account, but no card is required. That's enough to check whether the output quality works for you before paying anything.
- Pricing and funding. Subscriptions start at €29.90 a month, no commitment. Two certifications are state-recognized and therefore eligible for funding: RS7439 (content marketing with AI) and RS6792 (AI and sales). You can verify that they exist and what they cover directly on the official France Compétences registry. There's also a B2B track for training organizations that want to build these paths into their own offering.
This method has a limit, and it's worth saying out loud: if you're already an editor working on long projects with dozens of tracks, you'll stick with professional software. It's built for social video — short, produced at volume.
Method 2: consumer AI video editors
If you already have your footage and just want to cut it together, a standard AI video editor does the job, often for free for simple use.
These tools assume the raw material already exists. They automate assembly, not creation.
- CapCut: the most widely used for vertical formats. Automatic subtitles in French, background removal, smart reframing, a template library. The free version covers most needs; the advanced AI features and watermark-free export are paid.
- Descript: editing through text. The tool transcribes your video, you delete a word in the transcript and the matching shot disappears from the timeline. Brutally effective for interviews, filmed podcasts and tutorials. Interface in English.
- Adobe Premiere Pro: the professional benchmark, which now offers the same text-based editing principle. Powerful but dense, and the subscription sits in a different price bracket. The official documentation on text-based editing explains how it works in detail.
- Opus Clip and similar tools: you drop in a long video and the tool automatically pulls out short clips with subtitles and vertical reframing. Handy for recycling a livestream or a podcast. The waste rate is still high: out of ten suggested clips, two or three are genuinely publishable.
Quick way to choose: short formats and speed, go with CapCut; lots of talking to cut down, go with a text-based editor; recycling long-form content, go with a clip extraction tool. And if your end goal is to publish regularly, the next step is to automate publishing to social media — otherwise the time you saved editing gets swallowed by uploading.
Method 3: the automatic editor already sitting in your phone
Your phone already contains a free automatic editor, and for a simple video it's more than enough.
This is the option most articles forget to mention, even though it costs nothing and is already installed.
On iPhone, the Photos app offers “Memories,” assembled automatically from your videos and images, with music and transitions. iMovie, free on the App Store, adds “Magic Movie” mode: you select some clips and the app generates a paced edit you can then adjust shot by shot. On Android, Google Photos builds automatic videos around a theme or a person, and Samsung phones include an editor with automatic suggestions right in the Gallery.
What it does well: stringing clips together in chronological order, landing cuts on the music, adding a title, exporting straight to the right format. All in three minutes, no account to create, no credits to burn.
What it doesn't do: generate images that don't exist, produce a convincing voice-over, write subtitles in a custom style, or sustain a publishing rhythm of several videos a week. The moment you move from memory video to content video, you hit the ceiling fast. But for a first video, try this option before paying for a subscription: at least you'll know what else you actually need.
The 6-step workflow that works every time
A successful AI edit always follows the same order: write first, edit second, dress it up last. Flipping those steps is the number one cause of failed videos.
- Write the script before you film or generate anything. Three sentences is enough for a 30-second video: a hook, one useful point, one action to take. That text is what sets the final length — not the edit.
- Gather the raw material. Either you film it, or you generate the shots with AI, or you mix both. Always plan for 30% more footage than you think you'll use.
- Get it transcribed. The transcript is the backbone of modern editing: it feeds the subtitles, the cutting and the caption on your post.
- Cut. Remove the silences, the hesitations, and above all the first five seconds if they don't hook. A short video is won or lost in the first three seconds.
- Add the sound. Audio cleanup first, voice-over next if needed, music last and at low volume. If you're producing in French, timbre and accent are decisive: we compared the results in our test of French AI voice-overs.
- Dress it up and export. Readable subtitles (large, centered, two or three words at a time), one format per platform, and 1080p is plenty in almost every case.
Budget 20 to 30 minutes for a short video once the workflow is dialed in, versus two hours of classic manual editing. The real win isn't the speed of one video, it's being able to publish ten a week without spending your evenings on it. To apply this workflow to vertical format, our guide on creating Reels and TikToks with AI breaks down the settings platform by platform.
The mistakes that ruin an AI edit
Most disappointing AI videos fail for reasons of method, not technology.
- Trusting automatic subtitles without reading them back. Proper nouns, brand names and liaisons get transcribed badly. Thirty seconds of proofreading saves you from a nonsense line displayed in huge type on screen.
- Letting AI set the pacing. Automatic edits produce shots of uniform length, which feels mechanical. Vary it by hand: one long shot, two short ones.
- Piling on effects because they're there. Zooms, transitions, filters: every one added “because the tool offers it” pulls the viewer further from your message.
- Generating images before writing the copy. You end up with pretty shots that say nothing, and then spend an hour writing a script to fit the images.
- Editing once for every platform. A 16:9 format hastily cropped for TikTok chops off faces. Take the time to check the vertical framing.
- Neglecting the sound. A mediocre image with clean audio gets a pass. A beautiful image with dirty audio does not. It's the most consistent rule in video.
What to take away
AI doesn't edit your video for you: it removes the mechanical tasks so you can spend your time on the decisions that matter.
Start with the simplest thing: run your phone's automatic editor on an existing video to work out what you're missing. If what you're missing is raw material (shots, a voice, music), move to an environment that bundles those tools with credits included and a method to go with them. If what you're missing is just assembly speed, a consumer AI editor will do fine. In all three cases, publish your first video before hunting for the perfect tool: publishing is what teaches you, not comparing subscriptions.
