An edit can be technically correct and still remove an important part of the reasoning. That is why our workflow starts with a transcript and a cut plan that can be reviewed before rendering.
Separate source material from decisions
The original file is preserved. The transcript records words and timestamps. The plan describes the selected segments, titles and reasons. Each output gets its own version, so the edit can change without erasing what came before.
projeto/
project.json # editing decisions
raw/ # source media
data/ # transcript and timestamps
out/ # rendered versions
history/ # plan history
Explain each selection
This is a fictional example of the clips array used by our workflow. The timestamps only illustrate the format; for a real video, they come from the transcript and an audio check.
[
{
"id": "corte-01",
"start": 12.4,
"end": 48.8,
"title": "One idea per clip",
"why": "The segment includes a problem, an example and a conclusion."
}
]
The reason helps you review the selection. A good title needs to be supported by the segment. Pauses, repetitions and filler words should not disappear automatically if removing them changes the meaning.
Check it before calling it finished
The workflow combines technical checks with visual and audio review. Code can check duration, resolution and whether subtitles exist. Framing, synchronization and the meaning of the speech need to be evaluated in the final video.
The JSON above is an excerpt of our project’s data format, not a complete editor or a universal configuration file. Running the workflow requires the scripts, media tools and a local transcription model.
A clip can also become an article
After reviewing the transcript, extract the central question, the example and the conclusion. Rewrite it for reading and add any context that was implicit in the video. This creates an editorial starting point; it does not replace fact-checking or approval to publish.
The workflow was developed after analyzing Moisés Melo’s public guide ↗, with our own implementation and adaptations to our setup. The inspiration became a documented process, without treating available code as proof of an approved edit.
Draft for Davi Wallace’s review. The examples are for learning purposes. This article has not been published yet.