Editing takes longer than filming
Finding clean takes, cutting pauses and filler, timing captions and placing supporting clips consume the hours that should go to ideas.
PROJECTS / FIRST-HAND WORK
Build. Measure. Explain.
Phillipe, the AI video editor I co-founded, is my main project. Below it are earlier projects, each written up with the problem, the system, the evidence and the limits.
PROJECT 01 / PHILLIPE
Phillipe, an AI video editor for creators.
A product I co-founded with Sebastian Gonzalez Capdevielle for creators and personal brands who film themselves talking. It turns raw footage into a finished, designed short, vertical 9:16 or horizontal 16:9, that the person reviews, corrects in chat on a paid plan and approves.
Rough cut first, then style
Phillipe transcribes the speech with word-level timing, cuts pauses, filler and repeated takes, places B-roll on spoken beats, then adds word-timed captions, pacing and music.
Designed, not just captioned
Word-by-word kinetic typography, a hero word carrying each line, script and serif accents, text that can sit behind the speaker and B-roll cut on the spoken beat.
Starts free
The free plan includes one finished video a month. Paid plans add custom instructions and corrections in chat, and every edit shows a credit estimate before it runs.
See plans on phillipe.aiLive at phillipe.ai
Built for short videos of a person talking, vertical or horizontal, not for long-form productions. The person reviews the rough cut and approves the result. The Clean style is still preview-only.
Visit phillipe.aiRead why I built itHow the edit works
Step-by-step articles on the parts of the edit Phillipe automates.
Edit a talking-head videoWord-by-word captionsB-roll on the right beatEARLIER WORK / 02
AI assistant for Skool communities.
A temporary experiment that communities could invite, connect to their knowledge and use to support their members.
Knowledge becomes difficult to navigate
Members repeat questions, important answers disappear inside feeds and community managers spend time responding manually.
Ingest, answer and engage
The assistant learned community knowledge, answered member questions, created engagement posts and supported about-page testing.
Reached 50 communities
The project was used across 50 Skool communities. It was a temporary experiment and was never monetized.
Read the full case studyEARLIER WORK / 03
Automated LinkedIn research and publishing.
An anonymized vitamin company needed a more consistent way to understand competitor content and publish relevant company updates.
Competitor and company intelligence
The system researched competitor posts and received internal company context through Slack.
Two daily LinkedIn posts
It generated and published content informed by both outside patterns and the company’s actual activity.
Time saved and revenue lift
The client reported about four hours saved weekly and a three-percent revenue increase during the project period.
Read the evidence and limitationsALSO / ONE1 STUDIO
Visit One1 StudioWhat is slow, expensive or repeatedly breaking?
One1 Studio, the AI development studio behind Phillipe, also works with businesses: it starts with the process and designs the system around the result.