Three real photos of me went in.
A movie came out.
No camera, no crew, no acting lessons — Claude wrote every shot.
This is the whole trailer. Watch it first.
Every clip below was generated with Seedance 2.5 on Higgsfield, from my real photos — then run through a 35mm film-grade pass.
One at a time, full width. Click any poster to play.
Scene 1 — rooftop at dawn, drones rising behind me. Seedance 2.5 on Higgsfield, my real photos as the character reference, film-graded.
Scene 2 — the robot hangar salute. Same character sheet, same face, totally new set.
Scene 3 — full sprint through a neon canyon while a robot hands me coffee.
Scene 4 — the golden ENTER key slam. Look closely: it spells "ENER". AI still can't spell — the actor held, the typo didn't.
The movie-poster walk — explosion behind, no looking back.
The standup. A tiny robot works the mic, I heckle from the booth, it fires back — and wins. Seedance 2.5, film-graded.
The gym. My robot spotter keeps stacking absurd plates — and gives me a thumbs up while the bar bends. Seedance 2.5, film-graded.
The meeting. I slide one cookie across the boardroom table — six tiny robots pass it down the line like a ritual. Seedance 2.5, film-graded.
The dog walk. Golden hour, robot dog, one real squirrel — and I'm airborne. Seedance 2.5, film-graded.
The cooking show. The robot chef flambés way too hard, singes its own antenna — I extinguish it and look straight down the lens. Seedance 2.5, film-graded.
The robot lunch-lady toast — a tribute to Duncan's sci-fi love story.
This handheld talking-to-camera style is what brands pay creators for.
Now one character sheet makes it on demand.
UGC testimonial. Phone at arm's length, hoodie, real street — the "you have to try this app" clip. Brands pay creators for exactly this format; Seedance made it from three photos. Vertical 9:16, generated.
UGC desk react. Ring-light-at-home look, points at the monitor, gives the camera the "can you believe this" face. The reaction-ad staple — product demos, launch reactions, before/afters — now one prompt away. Vertical 9:16, generated.
Scene 4 literally has a typo on a giant golden key — I left it in on purpose.
The machine isn't perfect; it's consistent. Your face holds even when the spelling doesn't, and that's the part that used to be impossible.
Two layers: ask for film in the prompt, then finish it like film in the edit.
Steal the whole recipe.
Grain plus letterbox instantly read "film" instead of "AI render". The UGC clips skip the letterbox — phone footage isn't widescreen.
ByteDance's new video model landed on Higgsfield this week, and it is the first one that feels like a film camera instead of a slot machine.
The numbers are silly.
30-second clips, 50 reference images, 5,000-character prompts — 720p today, with 1080p and 4K on the way. That reference slot is the whole trick: it is how my real face got in.
Consistency — the same actor, the same outfit, shot after shot — and cinematography: real camera language like a 40mm large-format look, instead of screensaver drift.
Every film on this page came out of the same three-step machine.
One image that locks who you are — face, build, outfit — so every scene stars the same you.
You pitch Claude the movie in plain words, and it writes the shots like a director.
Seedance's superpower — one prompt with timestamps becomes a fully-cut sequence.
Tell Claude your age, build, hair, and outfit — and drop in a few real photos so it knows your face.
Then generate it on Higgsfield with GPT Image 2: neutral grey background, 2K minimum, 16:9.
Watch the step: a real Claude session, replayed at reading pace. The exact prompt goes in; Claude writes the image-generation prompt back — the gold line is the outfit lock you'll reuse in every scene.
It is a chat window. You paste one prompt into Claude, answer four questions about yourself, and copy its answer into Higgsfield.
If you can order a pizza online, you can make a character sheet.
Give it the premise, the vibe, and who's in it — Claude comes back with full scene prompts: location, wardrobe, props, dialogue, and camera.
Watch the step: a real Claude session, replayed at reading pace. Five ideas come back, Claude picks the winner, and every scene prompt carries the same gold consistency line — word for word.
The pitch that became the trailer at the top — recreated from the real session. You talk like a producer; Claude writes like a director.
I never even open a website — Claude sends the scene prompt through the Agent OS to Seedance 2.5, and the finished clip lands in my folder.
Watch the step — the real generation flow: Claude sends the prompt through the Agent OS to Seedance 2.5 (the render phase is time-lapsed; the real run took 5 min 03 s). The ending is the actual resulting clip.
No Agent OS? Do it manually: higgsfield.ai → Video → Seedance 2.5 → attach your sheet → paste the prompt → Generate.
Duncan's spaceship love story ran 15 shots from one prompt. Tag your uploads as @image1, @image2 — Seedance matches every reference to every beat.
You can wire this together yourself with the steps above. Or get the whole thing done inside the Agent Operating System — character sheets, scene prompts, and multi-shot sequences, pre-wired.
No — that's the biggest myth about it. The everyday 90% runs on a free local model on your own machine, free APIs slot in for more, and for the frontier work it drives the CLIs you already pay for — your Claude subscription already includes the Claude CLI, and the Agent OS plugs straight into it, so you're not paying twice.
And inside the AI Profit Boardroom there are full token-optimisation tutorials, so you cut usage to the bone and never think about it again.
Repeat your character description word-for-word in every prompt, and attach the same character sheet to every generation — never paraphrase, never swap the image.
Duncan's workflow goes way past one character — same sheet trick, three more toys.
Ask Claude for you at 60 and at 80, attach the same sheet — same outfit, same face, forty years later. Duncan aged himself a whole lifetime for one love story.
The character-sheet trick works on objects — Duncan built a tray of sci-fi cafeteria food that stayed identical across a 40-year story.
Describe the location, materials, and lens once — every character and prop inherits its look and lighting. Arguably the most important sheet of all.
I made one sheet — now every arm of the business gets cinematic footage of "me" without me ever standing in front of a camera.
Every video on this list used to need a camera, a setup, and an afternoon of takes. Now it needs a prompt.
Wrong: "I'm not a filmmaker — I wouldn't know where to point a camera."
Right: You never touch a camera. Claude writes the shot list, and "40mm large-format" is just words in a prompt.
Wrong: "AI video is too expensive to even play with."
Right: You start with one image and one 10-second scene — not a production budget. The whole trailer above came out of one afternoon.
Wrong: "My face will look weird — AI always melts faces."
Right: That was the old models. Real photos + the same sheet every time is exactly how my actual face held across all seven films on this page.
158 pages of members who already broke through these exact beliefs — real businesses, real wins, in their own words.
Read the 158-page wins doc →
Clear face, decent light — phone photos are fine. That's your whole "casting" process.
Give it your age, build, hair, and outfit, plus the photos — generate on Higgsfield with GPT Image 2, grey background, 2K, 16:9.
Pitch Claude one moment — you walking out of an explosion works — and feed the prompt plus your sheet to Seedance 2.5.
That's the whole machine at miniature scale. Once you've seen your own face in one cinematic shot, you won't stop.
This page gave you the whole AI Movie Machine — the sheets, the prompts, the sequence trick. The Boardroom is where you skip the assembly: the prompt pack pre-wired, the walkthrough on video, and coaching calls where we build your first film together.
Members are already using this exact stack for client videos, channel intros, and product films — across agencies, ecom, coaching, and SaaS in 38 countries.
Decide which one you are tonight.
Get the Agent OS → Inside the AI Profit Boardroom · skool.com/ai-profit-lab