My video machine
I built myself an assistant that edits the videos I never had the energy to touch.
2026. On Instagram, @meravish.s.
From a half minute recording, five moments made it in
Each stroke is a sentence that made it into the reel. Click or tap one to hear it.
From four and a half minutes of recording, four moments made it in
Each stroke is a sentence that made it into the reel. Click or tap one to hear it.
From a one minute recording, twelve moments made it in
Each stroke is a sentence that made it into the reel. Click or tap one to hear it.
My voice from one recording, over hand footage from July: my little daughter and me, on the same paper.
A creative guidance session with Liron, with the whole artwork in frame.
The time the assistant wrote my name without the yod. Hands, paper and a stream, with our chat on screen.
Months of footage waited on my phone
Hands on paper, painting sessions, voice recordings of me talking. Everything was filmed, and nothing was edited.
A good reel is not cut out of one window in a recording. It is stitched from moments: a sentence from here and a sentence from there, each one cut exactly at the end of a word, with captions that move with it. That is work of judgement, and it always got pushed back.
The assistant edits, and I decide
Five things it does for me, from raw footage to a finished reel:
- Light working copies. Every video gets a light copy to work on, and the original stays as it is.
- A transcript. Every word in the recording is written down with the exact moment it is said.
- A plan for each reel. Which sentences go in, from which recording, and what is on screen while they are heard.
- Captions. The word being said right now lights up in yellow.
- A check. At the end the assistant transcribes the finished reel again, to catch a word clipped at the edge.
In Hebrew, every word gets the font of its own script
The Hebrew caption font has no Latin letters, so a word like AI breaks in it. The Latin font of the same family draws Hebrew as circles. So the assistant picks a font for each word by its script, and every build checks a frame that has both languages.
The shadow was checked too. Three layers of shadow looked coarse on bright footage. Two soft layers read even on a white wall, without a black halo around the letters.
I packed the same machine for Liron
Liron and I are building PlaySoul together. She runs the same machine herself, on her own computer, from raw footage in the Drive to a checked reel.
Along the way it stops three times to ask her: is the transcript right, word for word? Are the chosen moments what she wanted to show? Should the finished reel go up to the Drive?
Something on your mind? Let us talk.
Fifteen minutes to explain what you need, and to see whether I can help. No deck and no commitment.