Phillipe is an AI video editor for short-form video: vertical for Instagram Reels, TikTok and YouTube Shorts, or horizontal for YouTube. You upload a clip of yourself talking, pick a style, and start the edit with one click. Phillipe transcribes the speech, cuts the pauses, filler and repeated takes, places your supporting footage on the moments it illustrates, and returns a rough cut for you to review. It then adds word-timed captions, pacing and music, and you ask for changes in chat until the video is right.
This case study starts with how Phillipe began, then explains why it was built the way it is. Its design decisions are the same ones One1 Studio applied to client work before it began building its own products: choose one narrow job, remove the step people struggle with, put human review where it is cheapest, and make the cost of each unit of work visible before it starts.
How Phillipe started.
Max Ferrer had been looking for a good AI video editor for a long time. He tried the tools. Most of them could cut a clip and put subtitles on it. None of them did what he actually needed:
- Masks, so text can sit behind the person on camera instead of floating on top.
- Text puzzling, where words of different sizes lock together into one designed block, with the important word carrying the line.
- Edits that look good and keep people watching, not a transcript pasted over a video.
What he wanted was an edit that looked like someone with taste had touched it. That was the part nobody had automated.
So he built one. It worked. It wasn’t good.
So Max decided to build it himself, and he did. The first version took his footage and gave back an edit. But it wasn’t that good. It was missing the thing that separates a video you scroll past from one you watch: an editor’s touch. Max isn’t a video editor. He could build the system, but he couldn’t teach it a craft he didn’t have.
Then he met Sebastian.
That’s when Max found Sebastian Gonzalez Capdevielle, a professional video editor who has edited videos for brands and creators like Whop and Clix, and told him what he was building. Sebastian was skeptical. He didn’t believe AI could edit like a professional, and from everything either of them had seen, that was a fair position.
They taught it like a student.
So they stopped treating the AI like a tool and started treating it like a student. Sebastian gave it instructions the way he would give them to one of his own students: what to cut, which word should carry a line, where text should sit, when a clip should land. They watched what came back, corrected it and tried again. They iterated, and worked hard at it, until they finally got a decent edit, the first one that looked like an editor had been there.
An editor with taste.
That’s when Max asked Sebastian to build it with him. Together they built Phillipe: an AI editor that doesn’t just edit videos. It edits them with taste. One1 Studio, the AI development studio Max founded, builds it.
The problem: AI editors didn’t make videos look good.
A video only works if people keep watching. The AI editors Max tried could cut a clip and put subtitles on it, which is a transcript pasted over a video, not an edit. None of them set text behind the person on camera, locked words of different sizes into one designed block, or made an edit that looked good and kept people watching. Most AI editors make a video shorter. Phillipe is built to make it look good.
The usual way to get more out of an AI tool is to type instructions into it, which asks for a different skill: describing an edit in words. A blank prompt box is a hard place to start if you do not already know what a good edit looks like.
So the goal for Phillipe was narrower than “AI for video”. Max and Sebastian wanted anyone to get Sebastian-level taste without writing a prompt: someone who films themselves talking should get an edit that looks good without editing and without prompting. The product’s tagline says it directly: “Great edits. No prompting.”
The approach: five decisions.
1. Narrow the job before building anything.
Phillipe does not try to edit every kind of video. It is built for one format: a person talking to camera, cut into a short video, vertical or horizontal, with optional supporting clips, photos and audio. That constraint made every later decision easier. The editor always knows the shape of the result, what a good cut looks like for that shape, and which mistakes matter most.
This is the step businesses most often skip. “Use AI for video” is not a project. “Turn the customer questions we answer on camera into short vertical clips every week” is.
2. Replace the prompt with a style.
Instead of asking for instructions, Phillipe asks you to pick a style. The style carries the craft: word-by-word kinetic typography, script and serif accents, text set behind the speaker and B-roll cut on the spoken beat, so the finished video looks designed rather than simply captioned. Kinetic, its signature style, gives every word a little motion, with sculpted typography, flowing script and expressive graphic moments. Cinematic leans into cinematic storytelling. A simpler style called Clean is shown as coming soon and is not available yet.
The style carries the creative direction, so an edit can start from a single click. On paid plans you can add a short note when you have something specific in mind; a free edit is Phillipe’s own first pass. The design choice was to make instructions optional rather than required, so the quality of the result does not depend on how well someone writes a prompt.
3. Ask people for what a machine would have to guess.
When you upload footage, you label each video as A-roll, the clip of you talking that carries the story, or B-roll, the supporting footage. Phillipe trusts that label instead of trying to infer it. The label takes a second to set, and it means the edit is built around the clip that actually carries the story rather than a guess about which one does.
The general lesson is simple. If a person can supply a fact with one click, ask for it. Inference is worth building only where asking is impractical.
4. Review the structure before the polish.
Phillipe works in two passes. The first is a rough cut: it transcribes the speech with word-level timing, removes pauses, filler and repeated takes while keeping the complete story, and places B-roll on specific spoken beats. You review that cut before anything else happens.
Only then does the style pass run: word-by-word captions timed to the speech, pacing, supporting footage and music, followed by the render. Putting the human checkpoint at the rough cut means the most important judgment, whether the story is right, is made before time and compute are spent on captions and effects. Fixing a story problem after the final render is the expensive way round.
5. Show the cost first, and leave the final call to the person.
Phillipe meters editing in credits and shows a credit estimate before every edit starts. The estimate is an estimate, not a guaranteed final charge, but it means nobody starts a job blind. After the first cut, you ask for changes in chat and make the final call yourself. Every correction is charged in full, in credits, so the cost of an edit reflects the work actually done.
When an edit needs a decision along the way, the question appears in the project chat. You answer it there, or continue if that option is offered, without creating a new project or uploading everything again.
What Phillipe does today.
- Upload your footage: video, photos and audio. Label each video as A-roll or B-roll.
- Pick a style: Kinetic or Cinematic.
- Optionally add a little direction, on a paid plan.
- Check the credit estimate and start the edit with one click.
- Review the rough cut.
- Get the styled version, with word-by-word captions, pacing, B-roll and music, as a vertical 9:16 or horizontal 16:9 video.
- Ask for changes in chat, and download when you are happy with it.
Two editors are available: Phillipe V1 on every plan, and Phillipe V1 Max, described as its most capable editor, from the Creator plan up. A free plan includes one finished video a month. Phillipe does not use your footage to train its own AI models, the site carries no advertising or cross-site tracking, and there is a connector for using Phillipe inside ChatGPT.
For a walkthrough of each screen, Phillipe’s articles include a tutorial on how to edit a talking-head video with Phillipe, step by step.
Who Phillipe is for.
Phillipe is for creators and personal brands who film themselves talking and want the finished video to look designed: type that moves with every word, text layered behind the speaker, B-roll that lands on the beat. Most AI editors stop at simple subtitles. Phillipe’s whole focus is how the edit looks and feels.
Phillipe is not a general video editor. It is built for short-form output from recorded footage. It does not generate footage from text or create avatars, and for long-form editing other tools are a better fit.
Lessons for businesses considering AI.
The same decisions apply well beyond video. They are the questions One1 Studio asked of every workflow when it implemented AI for businesses, and the ones it now asks of its own products:
- Is the job narrow enough? A defined input, a defined output and a shared standard for “good”.
- Where is the blank box? If people must write detailed instructions to get value, results will depend on who is good at writing them. Presets, templates and examples lower that bar.
- What can a person tell the system in one click? Labels, categories and approvals are cheaper than inference and easier to trust.
- Where is the cheapest review point? Put the human check before the expensive step, not after it.
- Is the unit cost visible? An estimate per unit of work makes a system easier to budget for and easier to trust.
- Who makes the final call? Automation should prepare the decision. A person should own it.
One lesson came from building Phillipe rather than from client work: taste can be taught, if the teacher has it. The breakthrough wasn’t a bigger model. It was a better teacher. Once a professional editor explained the craft the way he explains it to his students, the edits changed. If you are building an AI product for a craft, find the person who has the taste, and let them teach it.
If you want the wider context on where AI fits in a video workflow, Phillipe’s complete guide to editing videos with AI covers the process end to end.