the AI

How the AI helps, and where it stops.

On Light the AI is yours: Claude, Codex or Cursor, connected through the ReadyToEdit MCP. On Pro and Team, ReadyToEdit’s own AI also reads what you connect and writes the first draft from it. The decisions after that are yours. Below, where each one runs, and one place ReadyToEdit reads closely: the opening of a video.

Where it runs

Your AI on Light. Ours as well on Pro and Team.

Light: your own AI, over the MCP

Connect Claude, Codex or Cursor once. ReadyToEdit gives it your footage, scripts, audience and numbers, and turns the script it writes into a Footage preview and an editor link. ReadyToEdit runs no model of its own on Light.

  • Writes the script from your idea, your notes or your repo, and saves its lines in ReadyToEdit
  • Puts your clips on the lines, by file and range
  • Reads your numbers and the reference posts you saved
  • Writes to the audience, style and goals you set for the project
  • Hands you back the Footage preview and the editor link

The ReadyToEdit skills for Claude, Codex and Cursor give your AI the steps.

Pro and Team: ReadyToEdit’s own AI

Full plans

Everything Light has, and ReadyToEdit’s own AI reading what you connect and writing from it. Your own AI still works the same way.

  • The ReadyToEdit assistant, in Studio
  • Drafts and platform versions written from an idea
  • Footage found for each line of a script
  • Every recording read: search it by what is said or shown, and read a cut’s opening second by second
  • Library memory: a weekly Read of stories worth a video, and links between your work
  • Your voice, learned from your posts
  • What a project is about, read from its material
  • Formats proposed from what your material supports

Your opening, second by second.

Select the opening of a cut, up to 30 seconds, and it reads it closely: when you start speaking, when the picture first changes, how fast you talk, and what is written on screen.

What it reports

Each finding names the second it happened at, so you can scrub there and check it.

  • 0:00 – 0:00.40

    Nothing is said for the first 0.4 seconds of this span. The first word lands at 0:00.40.

  • 0:00.40 – 0:01.75

    You start speaking at 0:00.40, 1.35 seconds before the picture first changes at 0:01.75.

  • 0:00.40 – 0:03

    208 words per minute — 9 words between 0:00.40 and 0:03.

  • 0:03 – 0:04.20

    The span ends part-way through a line that runs on to 0:04.20: “and that is the part nobody warns you about”

What it compares against

Your own previous openings, and nothing else: no outside benchmark, no rule of thumb about hooks. Each line prints its count, so you can see how much it rests on.

  • 9 of your 12 other read hooks open with a question. This one opens with a finished statement.

  • The first word here lands 0.4 seconds in. 3 of your 12 other read hooks start speaking sooner than that.

  • You open this one at about 208 words a minute. Your 12 other read hooks run between 164 and 233, so this sits inside the range you already work in.

Some openings are written on screen.

It also reads any words on the picture back exactly as they are written, with when they appear and when they are gone.

Nothing is said out loud in the three seconds below. The whole opening is on the card.

0:00

The first sampled frame. Nothing to read yet.

0:00.25

We deleted a third of the codebase last quarter

One frame later the card is up, and stays up for ten.

0:02.75

On purpose

It resolves with two frames of the span left.

Fig. 1ExampleThree frames from the opening. The words on them match the lines the close read returns below.

What came back

  • Text on screen from 0:00.25 to 0:02.50, across 10 frames: “We deleted a third of the codebase last quarter”

  • Text on screen from 0:02.75 to 0:03, across 2 frames: “On purpose”

It will not tell you whether the hook is good.

There is no score, and none is planned. Boyd, Blackburn and Pennebaker tested structural conformance against popularity across 18,074 films and 639 novels and found nothing (Science Advances, 2020). Wu and colleagues found the rank correlation between views and watch time runs +0.80 in Music and −0.34 in News: when an outcome flips sign from one niche to the next, no threshold on an input carries over to yours. A score out of a hundred cannot be checked. Every line a close read returns carries a timestamp, so you can check it.

See it on your own work.

Connect the AI you already use on Light, or let ReadyToEdit read a recording’s opening, second by second, on a Full plan.