Founder on-screen steps out his front door in selfie mode and admits he is an AI replica video, made for SaaS, optimized for Instagram in 9:16 路 Voiced in English
An AI video platform that turns one selfie and a short voice sample into a photoreal replica of a founder, then films complete videos with that replica from nothing but a written script. The same clone can deliver an explainer, perform a testimonial style ad, read a launch announcement, and the platform handles the cutting, the captions, and the posting after that.
The buyer is a founder or a solo marketer who already accepts that short video is the cheapest distribution left and still has not shipped any. The blocker is never the idea. It is the camera, the light, the twelve takes, and the hour of editing that a thirty second clip quietly demands, all of it sitting outside the actual job of running the company.
A self recorded launch vlog converts for that buyer because it removes the objection while it makes the argument. The entire clip is product output, so there is nothing left to claim afterwards. A viewer who suspects the footage is generated has to sit with how ordinary the morning looks, and a viewer who never suspects it gets the pitch anyway.
The prompt
Vertical 9:16, 30 seconds, 12 jump cuts, one spoken line per cut. Handheld daily vlog look: DV 16mm texture, gentle film grain, softened sharpness, subtle halation around sunlight, low contrast, realistic skin tones, natural motion blur, small exposure shifts. Keep handheld shake, drifting composition, autofocus hunting, rushed reframing, uneven zoom, brief accidental cropping of the face. Bright sunny morning light throughout. It must read as real footage from someone's ordinary day, never a polished commercial. No music, only the voice and light ambient street and room sound.
CHARACTER: a charismatic European founder in his late twenties, light blue eyes, expressive eyebrows, medium brown hair pulled into a low bun with a few loose strands, short neatly trimmed beard and mustache, fair skin with real texture, lean athletic build. Plain black crew neck t-shirt, dark casual pants, clean sneakers, identical in every shot. No jewelry, no watch, no glasses. He speaks English with a light natural accent, warm and playful, real person pacing with small imperfections, never a radio voice. Lip sync must be exact.
SETTING: his apartment entrance, a tree lined sidewalk, a bright desk by a window, a couch corner, a small sunny street. Warm golden light, no other people anywhere, not in the background, not in windows, not in reflections. No visible camera, no visible setup. No readable text, no logos, no on screen UI. The product in the testimonial shot is a plain unbranded white bottle with no label.
SHOT 1 (3s, arm's length selfie, must open the video): he pushes his door open and steps into the morning sun, walking, calm confident smile. Line: "This is not [founder name]. I'm his AI replica."
SHOT 2 (2s, walking selfie): the frame drifts, he casually reframes himself, small grin. Line: "Made from one selfie."
SHOT 3 (2s, jump cut to a fixed angle, as if a new facecam recording just started): bright desk by the window, he drops into the chair and looks straight down the lens. Line: "Need an explainer? Watch."
SHOT 4 (4s, same angle, real teaching energy): he explains, counting on two fingers, clear engaging instructor delivery. Line: "You type your idea, I film the whole video for you."
SHOT 5 (2s, jump cut, new selfie recording): couch corner, he raises the camera fresh, casual energy. Line: "UGC ads? Watch this."
SHOT 6 (4s, handheld selfie, real UGC testimonial delivery): he holds the plain white bottle near his face, eyes wide, excited creator energy, then breaks character with a knowing smirk on the last word. Line: "Guys, one week with this and I'm obsessed. See?"
SHOT 7 (3s, walking selfie, back outside): he shrugs with his free hand, amused. Line: "And the founder? He filmed nothing."
SHOT 8 (2s, close handheld selfie): he snaps his fingers on the beat, direct eye contact. Line: "No camera. No reshoots. No editor."
SHOT 9 (3s, energetic walking selfie): sunlight flares softly behind him, confident stride. Line: "It auto-edits, trims, adds captions..."
SHOT 10 (3s, same walk, new cut): he ticks platforms off on his fingers, one per word. Line: "...and auto-posts everywhere."
SHOT 11 (1s, quick close selfie): eyebrows up, one beat. Line: "It even runs your ads."
SHOT 12 (2s, arm's length selfie ending): he slows, warm smile, small wave on the last word. Line: "[Your brand]. Go clone yourself."
CONTINUITY: same face, hair, outfit and build in every shot. No duplicated limbs, no broken hands, hands holding the camera or the bottle must look natural. Props stay consistent. Every angle change is a clean jump cut, as if a brand new recording just started. Never static, always moving or speaking.
After generation, add burned in captions in the editor: bold white sans serif, lower third, one keyword per line highlighted in yellow.Why this works
The first line is the whole strategy: "This is not [founder name]. I'm his AI replica." It spends the hook admitting the trick instead of hiding it, which is the only move that makes the next 28 seconds interesting. A viewer who was half scrolling now has a reason to stare at the face and look for the seam. The door opening into direct sunlight in the same beat gives the eye somewhere to go while the sentence lands.
The two demo blocks in the middle are the actual argument, and they are performed rather than described. At the desk he settles into the chair and teaches, counting on his fingers with instructor energy. On the couch he flips into creator mode, bottle near his face, excited delivery, then breaks character with a smirk on the last word. Same face, two completely different registers, three seconds apart. That contrast proves range in a way no feature list can, and the smirk is what stops the testimonial reading as a lie.
The rest is built for retention mechanics. Twelve cuts in thirty seconds means a new frame every 2.5 seconds, so there is no dead moment to scroll on, and the cuts are motivated by the script rather than sprinkled over it. Every location is somewhere an ordinary person actually films: a doorway, a sidewalk, a desk, a couch. The objection block, "no camera, no reshoots, no editor," arrives at seventeen seconds, exactly when a curious viewer starts asking what it would cost them. Burned in captions carry it for the sound off majority, and the close lands on an instruction rather than a URL.
Ship one like this
Start with this prompt, swap your product in. Two minutes from idea to download.
Last updated August 2026