Where AI helps a UGC creator, where it does not, and why brands in Brazil still pay for a real face and a real voice.
Where AI really helps a creator
Six uses that save time without touching what the brand buys.
Breaking a brief down into a shooting plan, which avoids back and forth.
Proposing several hooks from the same argument, then sorting them by ear.
Generating subtitles, then correcting them, which halves the finishing time.
Cleaning the sound of an average take, when reshooting is not possible.
Translating a delivery message or a quote for a foreign client.
And sorting, naming and finding your own files, the invisible task that takes the most time.
Where AI does not replace you
Five things a tool does not supply.
The product in the hand, its real size, its texture, the sound it makes.
Proof of use: an identifiable person showing what they did.
Responsibility: somebody has to answer for what is claimed in the video.
The unexpected that makes a take credible, the hesitation, the failed gesture then repeated.
And the right to use a face and a voice, which is granted in writing and paid for.
What Brazilian brands really buy
Four elements, in the order in which they count.
A real, identifiable face, with an accent and vocabulary matching the target audience.
Proof that the person used the product, which no generation supplies.
A clear licence, which supposes a person able to grant it.
And regularity of delivery, which rests on an organisation, not on a tool.
One direct consequence: campaigns that replaced creators with generated visuals most often come back on product pages and retargeting, where the objection is lifted by proof, not by aesthetics.
The skills that raise your rate
Six skills, ranked by effect on price.
Writing a hook that holds for three seconds, tested and reworked.
Filming cleanly in natural light, with no extra equipment.
Editing fast, with cutaways prepared at the shoot.
Delivering several variants of the same video, which brands buy for advertising.
Reading a brief and asking two useful questions before shooting.
And invoicing properly, with a separate rights line.
The rules that protect you
Five rules applicable from today.
Declaring what was assisted, dealt with above.
Not using a tool that reproduces another person's face or voice, including for a demonstration.
Not generating a product result that has not been observed, particularly in regulated categories like health, food or finance.
Checking with the client whether their brief forbids AI, before shooting and not at delivery.
And keeping the original rushes, which prove the take is real.
Where the tool helps, and where it hinders
| Task | The tool helps | The tool hinders |
|---|---|---|
| Generating text variants | Yes, a real time gain | No |
| Subtitling and cutting | Yes, with a review | No |
| Fabricating a talking face | No | Yes, credibility lost |
What stays irreplaceable
The hand holding the object, the hesitation before saying something, the room's noise. Those details are exactly what the brand buys, and they are not generated.
What is disclosed to the audience
Any generated image or voice must be flagged. The exact duties change and are confirmed with the competent authority and a lawyer before each advertising release.
A workflow that keeps you fast and human
Six steps, half a day per video.
Read the brief and ask the questions the same day.
Write the script with assistance, then say it out loud and cut what does not come out naturally.
Prepare the cutaways before switching on the camera.
Shoot the a-roll, then the b-roll, in that order.
Edit, generate the subtitles, correct them by hand.
And deliver with the message stating the format, the rights covered and what AI touched.
One last marker: the tool moves the time, it does not move the value. What you sell remains proof of use attributed to a person, and that is precisely what a model cannot produce in your place.
