Antoni OS · Antoni Wijaya

Prompt dan tutorial AI siap salin.

Tutorial AI yang sudah saya coba dan buat ulang, dengan semua prompt siap disalin.

Video AI

Fan-blade fit check: reel ganti outfit 15 detik untuk pria

Lima outfit, satu bidikan dari atas, dan setiap pergantian tersembunyi di balik kipas angin langit-langit yang berputar. ChatGPT membuat gambarnya, Higgsfield membuat videonya.

Versi pria, dibuat ulang dari tutorial oleh @quinqueen

Yang Anda butuhkan

  • 2 foto diri Anda yang jelas, tampak depan dan samping
  • 5 referensi outfit seluruh badan, termasuk sepatu (Pinterest cocok)
  • ChatGPT dengan fitur pembuatan gambar
  • Higgsfield yang terhubung di ChatGPT, dengan model Seedance 2.5

Langkah 1: Buat character sheet Anda

Unggah dua foto Anda, isi tinggi dan berat badan, lalu kirim prompt ini. Hasilnya mengunci wajah dan postur Anda dari empat sudut.

Use the two photos I uploaded (front and profile) as the only reference for who I am. Keep my face, jawline, skin tone, hairline and hairstyle exactly as they are.

I am [___] cm tall and weigh [___] kg. Build my body to match those numbers naturally, without making me more muscular or slimmer.

Create one photorealistic character sheet on a plain white background, split into four panels: front headshot (top left), profile headshot (bottom left), front full body (centre), side full body (right).

Same man in every panel, wearing the same simple off-white loungewear, barefoot, standing relaxed. No text, logos, props or other people. Height, build and proportions must match across all four panels.

Setiap panel harus terlihat seperti Anda, terutama mata, hidung, rahang, dan garis rambut. Perbaiki di sini dulu sebelum lanjut.

Langkah 2: Buat lima outfit, satu per satu

Unggah character sheet Anda dan satu referensi outfit setiap kali, lalu kirim prompt ini lima kali.

Image 1 is me. Image 2 is an outfit reference. Dress me in the outfit from image 2.

Keep from image 1: my face, features, skin tone, age, build, and my current hair. Do not copy the reference model's hairstyle.

Take from image 2 only: the clothing (cut, fit, fabric, colour, pattern, layering), the shoes, and visible accessories such as a watch, belt, glasses, bag, chain or cap.

Ignore from image 2: the model's face, expression, pose, attitude, mood, lighting and photo style.

If the reference shows the outfit from several angles or in close-ups, use all of them to understand the details of the one outfit.

Full body, photorealistic, neutral studio background, natural relaxed stance, shoes clearly visible. I should look the way I normally do, just wearing these clothes, not like I turned into the model.

Avoid: a different face, plastic or over-retouched skin, a cold or overly high-fashion expression, distorted hands, wrong or missing accessories, clothing details that don't match the reference.

Beri nama Outfit 1 sampai Outfit 5, sesuai urutan yang Anda mau di video. Pastikan sepatu dan aksesori utama terlihat.

Langkah 3: Buat videonya

Di ChatGPT, unggah kelima outfit sesuai urutan, tandai Higgsfield, lalu kirim prompt ini.

Using the five images I uploaded, in upload order, create a photorealistic 15-second vertical (9:16) fashion video of one man changing through five outfits.

Identity and outfits
All five images show the same man. Keep his face, skin tone, build and hairstyle identical for the whole video. Image 1 is outfit 1, image 2 is outfit 2, and so on to image 5. Recreate each outfit faithfully, including shoes and accessories. Use the images only for identity and clothing; build the scene below from scratch.

Camera
One continuous shot. The camera is locked, pointing straight down at 90 degrees through a ceiling fan at the man lying on the floor below. No zoom, pan, tilt, orbit or angle change.

Room
Wall-to-wall carpet in a muted, dusty blue. A curved, pale pink sofa along the left edge of the frame, partly cut off. A small silver remote lies on the carpet beside his right hand. Nothing else in the room.

Fan
A large ceiling fan in polished chrome, with wide, brushed-metal blades, hangs very close to the lens, between the camera and the man. Its motor housing sits at the top centre, partly out of frame. The blades are wide enough to cover his whole body as they pass. The fan is still at first, and only starts turning counter-clockwise after he presses the remote. Keep the blade shapes readable, with light, natural motion blur.

Timing
0.0–1.8s: Outfit 1. He lies curled on his side facing the sofa, knees bent, as if dozing, then rolls onto his back: head toward the top of the frame, feet toward the bottom, arms loose at his sides.
1.8–4.0s: He picks up the remote beside his right hand, points it at the fan and presses once.
4.0–5.2s: The fan starts turning. A blade sweeps across his body and hides the first change.
5.2–7.8s: Outfit 2. Hands resting on his hips, looking straight up at the camera.
7.8–10.4s: Outfit 3. One hand in his trouser pocket, the other arm relaxed, a calm look up at the camera.
10.4–12.8s: Outfit 4. Arms loose at his sides, chin slightly lifted, confident.
12.8–15.0s: Outfit 5. One hand on his stomach, the other arm out on the carpet, head turned toward the sofa with a slight smile. End clearly on outfit 5.

Transitions
Exactly five outfits and four changes. Every outfit and pose change happens while a blade passes over him: where the blade hasn't reached yet, he still wears the old outfit; behind its trailing edge, the new one. No morphing, dissolves, flashes or uncovered jump cuts. The room, the camera, his size and his position on the carpet stay the same throughout.

Look
Natural movement and anatomy, believable fabric, soft, even daylight, consistent shadows. No duplicate people, extra limbs, face changes, floating body parts or extra cuts.

Use the Seedance 2.5 model.

Sebelum Anda unduh, cek

  • Wajah yang sama di kelima outfit
  • Urutan outfit benar, tepat empat kali pergantian
  • Setiap pergantian tertutup baling-baling kipas
  • Kamera tidak bergerak, dan seluruh badan tetap di dalam frame
  • Video berakhir jelas di outfit 5

Gambar AI

Bikin manga sendiri dengan AI (Higgsfield Mangaka)

Dari satu ide jadi komik 33 halaman. Setiap langkah dengan screenshot asli, dan komiknya bisa langsung dibaca.