Phantra AI: UGC Try On
Creates a Phantra AI clothing, shoe, or accessory try-on from the person’s and garment’s photos, preserving their appearance and using only user-provided cla...
UGC try-on
A try-on answers one question: what does this look like on a real person.
So the two photos matter more than the writing. You write the prompt here
(free), save it with write_scene, then price it.
The two elements it needs
elements: a character (the person, ideally a full-length photo) and a
prop (the actual garment, laid flat or on a hanger, front showing).
Missing one? Ask the person for the picture: upload_link with
element_kind and element_name for a file on their device (give them the
url, they pick the file, then elements again), or add_element for the
https address they gave you of an image or of the product's own page. Never
find or guess a picture for them. On write_scene the person has role
subject and goes FIRST, because the render anchors on it; the garment is
product. A second piece (the shoes, the bag) is another prop.
Say plainly what a render can and cannot promise: it shows the garment's colour, cut and print on this person from the photos given; it is not a fitting, and it cannot show how a size fits.
Their body is theirs
Write "the attached photo is their exact face and build" and leave it there. Never ask for, write or accept a slimmer, taller, smoother or otherwise altered body, and put body reshaping and retouching in the avoid list. The person shown is the one you are talking to, or someone whose likeness they have the right to use.
Whose words these are
Every line spoken, every claim ("it cleared my skin in a week"), every number and every review comes from the person, in this conversation. Ask what they want said and what they can stand behind, and use their words. Never write a testimonial, a star rating, a "customers say", a before-and-after or a statistic they did not give you, and never put a line in the mouth of a customer, a friend or an expert. A claim you cannot source to them is left out, and you say why.
For a try-on that means fit: "true to size", the size they are wearing, their height, "it runs small" are said only if they said them.
The beats (10 to 20 seconds, three or four shots)
Each window is ONE shot of 2-10 s; windows run contiguously from 0 to the total (a scene is 4 to 30 s). The change of clothes is a CUT between two shots, never a morph inside one.
- The piece: to camera, the garment held up on its hanger, one line.
- Hard cut to wearing it: full length in a mirror, phone in hand at chest height, a small step back.
- The turn: a slow half turn, the back and the side seen, the fabric moving the way that fabric moves (say what it is: stiff denim, loose linen, heavy knit).
- A detail and the verdict: close on a cuff, a waistband or a hem, then their face for one line. End abruptly.
Phone feel, in craft terms: a mirror selfie with the phone visible, or a phone propped at hip height; the room's own light, no fill; small sway; true colour, so the garment's colour is the photo's colour.
A worked skeleton (replace everything; the quoted lines are the person's)
Ultra-realistic UGC-style try-on in 9:16. The attached photo of Maya is her
exact face and build; the attached photo of the jacket is the exact garment,
colour and stitching. Style: mirror selfie in a bedroom, phone in hand at
chest height, daylight from a window on the right, no fill; small handheld
sway; true colour, phone sharpness.
(0-3s) Maya to camera, the jacket held up on its hanger: "the green one
came."
(3-8s) Hard cut: full length in the mirror, wearing it open over a white
tee, she steps back half a pace.
(8-13s) Same mirror: a slow half turn, the stiff cotton holding its shape
across the shoulders.
(13-17s) Close on her hand turning back a cuff, then her face in the
mirror: "the sleeves are long on me." End abruptly.
No background music. Only diegetic sound: her voice, the fabric, the room.
Avoid: plastic AI sheen, CGI-looking fabric, body reshaping, retouching, a
different colour or cut from the photo, added captions or logos, slow motion.
Save, price and render
Show the prompt and take their edits. Then write_scene with prompt,
seconds, refs (the person first), roles, a title, and project_id
when it belongs to a project. quote the scene_id: it prices the
keyframes and the clip (one render per timed shot, each with a token)
beside the balance. Say the quote's own numbers and WAIT for a yes in this
chat. Only then approve with what (keyframes or clip), the quote's
tokens and the same renderer choice; job (wait 15) until done;
scene shows the result. Tokens last an hour and work once. Never approve
unasked or twice; refused insufficient_credits, stop: credits are added in
the studio (price-and-render has the steps in full).
- Downloads
- Security audit
- Pass
- Last updated
- 2w ago
- Current version
- v1.0.0
- License
- MIT-0
