A believable, single-take testimonial delivered straight to camera by an AI avatar.
What this prompt does
Generates a natural, unscripted-feeling talking-head testimonial — the workhorse format for cold-outreach spec demos and always-on retainer content.
Best for
Skincare, wellness, supplements, personal care — any category where trust is the primary conversion lever.
Example render coming soon — this prompt is actively used in client work; a sample clip will be added here once we can share one.
▸ THE FULL PROMPT
handheld selfie-style talking-head video, one person speaking directly to camera in a bright, slightly cluttered home setting (natural light from a window, everyday background — not a studio)
subject: mid-20s to early-30s, casual clothing, relaxed posture, holding the product naturally in one hand, occasionally glancing down at it while talking
performance: warm, conversational, unscripted energy — mid-sentence pauses, a small laugh, natural hand gestures, direct eye contact with the lens broken only to glance at the product
camera: static handheld selfie framing, slight natural wobble, no gimbal-smooth stabilization, front camera aspect ratio 9:16
(0-3s) subject picks up the product off-frame, brings it into shot, opens with a hook line delivered mid-motion, not posed
(3-10s) walks through the specific problem the product solves in their own words, casual asides, product held up and gestured at naturally
(10-18s) the "moment of proof" — a visible reaction, a specific detail or result mentioned by name, genuine enthusiasm not overacted
(18-22s) closes with a simple, low-pressure recommendation, sets product down or holds it toward camera
lighting: soft natural window light, slight underexposure acceptable, no studio lighting rig visible
audio: room tone present, no music bed, natural background ambience (traffic, a fan, a dog once) at low volume
grain: light phone-camera compression artifacts, no cinematic color grade — authenticity over polish, no distortion on the face
Unlock the full prompt
Enter your email to reveal the complete copy-ready prompt and unlock the copy button. Free. Roster access.
Write the four spoken lines before you touch the prompt: a hook of under eight words, the problem in the customer's own words, one specific result, and a soft recommendation. Drop them into the timed beats. The model handles the performance, not the copywriting.
Take the problem line from real reviews or support emails, not from your brand copy. The phrasing customers use is what makes the delivery sound unscripted.
Give the model a photo of the real pack as a reference so the label reads correctly when it is held up between 3 and 10 seconds.
Render three to five takes and choose on delivery, not looks. The pause and the small laugh land differently every time.
Where it goes wrong
It sounds like an ad read. Cut the script by a third. Long sentences kill the mid-thought feel the prompt is built on.
The pack changes between shots. Keep the product in hand and in frame, and re-render rather than trying to patch it in the edit.
The avatar is presented as a real customer. That is a fake testimonial in the eyes of most ad regulators. Use the avatar to perform words from a genuine review, or label it.
What to test
Three hooks against the same body: a question, a blunt claim and a confession.
A proof line with a number against one with a sensory detail.
The full 22 seconds against a 12-second cut that keeps only the hook and the proof.