# Verbal Consent — On-Camera Script

Record this **at the start of every take**, before the real recording begins. It takes twenty
seconds and it is the strongest evidence you can hold: a person, on camera, in their own
voice, agreeing. It backs up the paper form and survives a lost or smudged signature.

Keep it as its own short clip, named with the same Asset ID plus `-CONSENT`.

---

## Crew says (camera rolling, subject in frame)

> "Today is the **first of September, two thousand and twenty-six**. We are at
> **[location]**. This is asset **[read the Asset ID aloud]**."
>
> "Please say your name."
>
> "Do you agree that this video and your voice can be sold to companies that build
> artificial intelligence, and used to train computer models?"
>
> "Do you understand that you are being paid once, now, and not each time it is used?"
>
> "Have you been paid?"

## Subject answers

Let them answer in their own words, in their own language. Do not coach the wording. A
natural "yes, I agree, they have paid me" in Yoruba or Pidgin is worth more than a recited
sentence in English.

---

## Rules

- **Face visible.** The consent clip must show the same face that appears in the footage.
- **Say the Asset ID out loud.** This is what ties the consent to the file. Do not skip it.
- **Record it even if the form is signed.** Both, always.
- **If someone hesitates, stop.** Do not film them. Move on. One unhappy subject costs more
  than one hour of footage is worth.
- **Under 18: do not film.** No exceptions on this trip. Guardian consent is a separate
  process and you do not have time to do it properly before Wednesday.
- **Bystanders.** If a face is identifiable in the background and unreleased, either blur it
  in post or cut the clip. Log the decision in the manifest `notes` column.