Demo ad · fictional product · 12.5 seconds, silent
Fashion · step-by-step tutorial
Silk Lounge Set AI ad with Wu Sefei
A loungewear designer sells a silk set by touch: fabric in the light, silk against her cheek, one slow turn. Below is every step we took, in order: which pack files we uploaded, the exact messages we sent to the AI, the settings, what came back, which takes we threw away, and how we cut the ad. Download the pack and you can make the same ad.
- Character
- Wu Sefei
- Product
- SOFT HOURS Silk Lounge Set (fictional)
- Length
- 3 shots + product end card, 12.5s
- Generations
- 4 images, 3 videos
What you need
- 1
Wu Sefei's character pack
Free with an account. Unzip it; the folders are numbered (01_character_reference, 02_turnaround_views …).
- 2
An image model that takes several reference images
We used GPT image model, image-to-image through the KIE API. Any image model that accepts 2–4 reference images plus a text message works the same way.
- 3
Kling 3.0 image-to-video
We used Kling 3.0 image-to-video (720p, no sound) with a character element (a saved set of reference images). Other image-to-video models work too; results will differ.
- 4
A video editor
CapCut, DaVinci Resolve, Premiere or ffmpeg. You only trim, add captions, add an end card and music.
Step 1Unzip the pack and find these files
The ad uses 8 of the pack's images. Nothing else was uploaded except the product still you make in step 3. Paths are exactly as they appear in the zip.
01_character_reference/full_body.pngUsed for: Product still, Character reference (video), Shot 2 first frame, Shot 3 first frame
04_props/prop_role.pngUsed for: Product still, Shot 1 first frame
01_character_reference/artwork.pngUsed for: Character reference (video)
07_expressions_poses/expression_smile.pngUsed for: Character reference (video), Shot 2 first frame
02_turnaround_views/view_front.pngUsed for: Character reference (video)
05_scenes/scene_workspace.pngUsed for: Shot 1 first frame, Shot 2 first frame, Shot 3 first frame
03_outfits/outfit_work.pngUsed for: Shot 1 first frame
02_turnaround_views/view_side.pngUsed for: Shot 3 first frame
Step 2Describe the character once
Open 01_character_reference/model_sheet.png and write down what must never change: skin, eyes, hair, outfit and art style. This paragraph goes, word for word, at the start of every message in the steps below (they already include it, so you can copy them as they are). For the video model the opening words become @wu is the same …, which points at the character reference from step 4.
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
Same adult woman as the reference images: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references.
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
She wears the champagne silk camisole and wide-leg trousers with the dusty-rose robe, never the ivory coat.
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
no extra people, no text or logos, no outfit change, no ivory coat, no scene cuts, no camera cuts, correct hands with five fingers
Step 3Make the product still
Make one picture of the product first. You reuse it for the end card, so the product looks the same everywhere. Never ask the model to write the brand name; add text in the edit.
You · Product still
1. Attach 2 images
Full body01_character_reference/full_body.png
Prop04_props/prop_role.png
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
Soft anime illustration product shot in exactly the same art style as the references. A champagne silk loungewear set, the same camisole and wide-leg tie-waist trousers as in the first reference, neatly folded on a pale marble table beside a folded dusty-rose silk robe, warm morning window light, soft shadows, a sprig of white flowers. Generous empty space on the left for a title. 16:9 still. No people, no text, no logos.
3. Settings
- GPT image model, image-to-image
- 16:9
- 1K
AI returns

- Same art style as the pack.
- No text, logos or people in the image.
- Empty space on the left for the brand name.
Step 4Save the character reference for video
In Kling 3.0 we saved these 4 images as one character element named wu with the description below. Every video message then starts with @wu, and Kling keeps the face and outfit from these images. If your video tool has no such feature, skip this step: the first frame and the character description still carry most of the look.
Full body01_character_reference/full_body.png
Artwork01_character_reference/artwork.png
Expression (smile)07_expressions_poses/expression_smile.png
Front view02_turnaround_views/view_front.png
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
wu
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
adult woman loungewear designer, champagne silk set and dusty-rose robe
Step 5Shot 1: Silk in the light
3s in the ad · caption: “Cut from washable silk”
1a · Make the first frame
You · First frame
1. Attach 3 images
Prop04_props/prop_role.png
Scene05_scenes/scene_workspace.png
Outfit03_outfits/outfit_work.png
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
Same adult woman as the reference images: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Close-up on the sunlit atelier table: her hand with a pearl bracelet lifts a corner of champagne silk from the open swatch book, the fabric catching warm window light. Only her hand and forearm in the dusty-rose robe sleeve are visible. Create one 16:9 cinematic still frame for this shot. She wears the champagne silk camisole and wide-leg trousers with the dusty-rose robe, never the ivory coat. Clean anatomy, no text, no logos.
3. Settings
- GPT image model, image-to-image
- 16:9
- 1K
AI returns

- Face, hair and outfit match the turnaround sheet.
- Hands have five fingers; no text or logos.
- If anything is off, generate again before making the video. The video can only be as good as this frame.
1b · Turn it into video
You · Video
1. Attach 1 image
Start frame: the image from 1aplus the @wu element from step 4
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
@wu is the same adult woman: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Close-up on the atelier table in warm window light. Her hand slowly lifts a corner of champagne silk and lets it slide off her fingers, the fabric rippling and shimmering. Static camera, very slow push-in, single continuous shot. Negative: no extra people, no text or logos, no outfit change, no ivory coat, no scene cuts, no camera cuts, correct hands with five fingers.
3. Settings
- Kling 3.0 image-to-video
- Standard mode, 720p
- 3 seconds
- Sound off
- Single shot (multi-shot off)
AI returns
1c · Check the take and trim it
- The first take was usable.
- Watch the whole clip. If the model cut to another camera angle, changed the outfit or added a second person, generate it again; this happens and is normal.
- Use 0–3 s of the 3-second clip.
- Save it as
shot-1.mp4.
Step 6Shot 2: Soft to the touch
3s in the ad · caption: “Soft enough to forget it's there”
2a · Make the first frame
You · First frame
1. Attach 3 images
Expression (smile)07_expressions_poses/expression_smile.png
Full body01_character_reference/full_body.png
Scene05_scenes/scene_workspace.png
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
Same adult woman as the reference images: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Chest-up shot in the sunlit atelier: she holds a fold of the champagne silk against her cheek with both hands, eyes closed, a small contented smile. Blurred clothing racks and flowers behind, warm light. Create one 16:9 cinematic still frame for this shot. She wears the champagne silk camisole and wide-leg trousers with the dusty-rose robe, never the ivory coat. Clean anatomy, no text, no logos.
3. Settings
- GPT image model, image-to-image
- 16:9
- 1K
AI returns

- Face, hair and outfit match the turnaround sheet.
- Hands have five fingers; no text or logos.
- If anything is off, generate again before making the video. The video can only be as good as this frame.
2b · Turn it into video
You · Video
1. Attach 1 image
Start frame: the image from 2aplus the @wu element from step 4
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
@wu is the same adult woman: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Chest-up shot in a sunlit atelier. With eyes closed she presses the champagne silk softly to her cheek and smiles; her curls move slightly in the light breeze. Static camera with a gentle push-in, shallow depth of field, single continuous shot. Negative: no extra people, no text or logos, no outfit change, no ivory coat, no scene cuts, no camera cuts, correct hands with five fingers.
3. Settings
- Kling 3.0 image-to-video
- Standard mode, 720p
- 3 seconds
- Sound off
- Single shot (multi-shot off)
AI returns
2c · Check the take and trim it
- The first take was usable(an earlier request failed to start and was not charged).
- Watch the whole clip. If the model cut to another camera angle, changed the outfit or added a second person, generate it again; this happens and is normal.
- Use 0–3 s of the 3-second clip.
- Save it as
shot-2.mp4.
Step 7Shot 3: The turn
3s in the ad · caption: “From studio to sofa”
3a · Make the first frame
You · First frame
1. Attach 3 images
Full body01_character_reference/full_body.png
Side view02_turnaround_views/view_side.png
Scene05_scenes/scene_workspace.png
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
Same adult woman as the reference images: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Full-length shot in the bright atelier: she stands mid-turn on the pale rug wearing the champagne silk set with the dusty-rose robe open, robe hem starting to swing out, warm window light behind her. Create one 16:9 cinematic still frame for this shot. She wears the champagne silk camisole and wide-leg trousers with the dusty-rose robe, never the ivory coat. Clean anatomy, no text, no logos.
3. Settings
- GPT image model, image-to-image
- 16:9
- 1K
AI returns

- Face, hair and outfit match the turnaround sheet.
- Hands have five fingers; no text or logos.
- If anything is off, generate again before making the video. The video can only be as good as this frame.
3b · Turn it into video
You · Video
1. Attach 1 image
Start frame: the image from 3aplus the @wu element from step 4
2. Send this message, word for word
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
@wu is the same adult woman: warm fair skin, brown eyes, long chestnut side-parted curls, pearl necklace and pearl stud earrings, champagne silk camisole, champagne wide-leg silk trousers with a tie waist, long dusty-rose silk robe worn open. Soft anime illustration style exactly like the references. Full-length shot in a bright atelier. She makes one slow graceful half-turn, the dusty-rose robe and wide silk trousers flowing outward, then she looks back over her shoulder with a soft smile. Static camera, single continuous shot. Negative: no extra people, no text or logos, no outfit change, no ivory coat, no scene cuts, no camera cuts, correct hands with five fingers.
3. Settings
- Kling 3.0 image-to-video
- Standard mode, 720p
- 3 seconds
- Sound off
- Single shot (multi-shot off)
AI returns
3c · Check the take and trim it
- The first take was usable.
- Watch the whole clip. If the model cut to another camera angle, changed the outfit or added a second person, generate it again; this happens and is normal.
- Use 0–3 s of the 3-second clip.
- Save it as
shot-3.mp4.
Step 8Cut the ad
Make a 1920×1080 project at 24 fps and lay the parts end to end with hard cuts (no transitions). The finished ad is 12.5 seconds.
| # | File | Use | Length | Caption |
|---|---|---|---|---|
| 1 | shot-1.mp4 | 0–3 s | 3 s | Cut from washable silk |
| 2 | shot-2.mp4 | 0–3 s | 3 s | Soft enough to forget it's there |
| 3 | shot-3.mp4 | 0–3 s | 3 s | From studio to sofa |
| 4 | product.png | End card | 3.5 s | SOFT HOURS / Silk Lounge Set / Made for slow mornings. |
Captions
- Bottom left: 110 px from the left, 170 px from the bottom.
- White, bold italic, 54 px (we used Avenir Next).
- On a rounded black box at about 47% opacity.
- Our “Demo ad · fictional product” label is only there because the products are invented; leave it out of yours.
End card
- product.png for 3.5 s with a slow push-in from 100% to about 110%.
- A cream gradient over the left 60% so the text reads.
- SOFT HOURS in bold italic, 94 px; Silk Lounge Set at 52 px; the tagline at 40 px in #b05c68.
- The text fades in 0.3 s after the card starts, over 0.4 s.
Music and export
- One track under the whole ad: fade in over 0.5 s, fade out over the last 1.5 s. Use music you have the rights to.
- Export H.264 MP4, 1920×1080.
- The Kling clips are 720p; scaling them to 1080p in the edit is fine for social video.
Prefer the command line? The same edit as an ffmpeg script
Put your clips, product.png, music.mp3 and two font files in one folder and run this. We ran it on our own clips and it reproduces the ad above.
The English text below is the exact prompt used. Copy it unchanged to reproduce the workflow.
#!/bin/sh # Cut the SOFT HOURS ad: 3 shots + a 3.5 s product end card, 1920x1080, 24 fps. # Put these files in one folder first: # shot-1.mp4 the video you kept for "Silk in the light" # shot-2.mp4 the video you kept for "Soft to the touch" # shot-3.mp4 the video you kept for "The turn" # product.png the product still # music.mp3 a track you have the rights to use # title.ttf font for captions and the brand (we used Avenir Next Bold Italic) # text.ttf font for the product name and tagline (we used Avenir Next Bold) # Needs an ffmpeg build with the drawtext filter: `ffmpeg -filters | grep drawtext` # must print a line (the static builds linked from ffmpeg.org have it). set -e # 1. Trim each shot, scale it to 1080p and burn in its caption. ffmpeg -y -ss 0 -t 3 -i shot-1.mp4 -an \ -vf "scale=1920:1080:flags=lanczos,fps=24,setsar=1,format=yuv420p,drawtext=fontfile=title.ttf:text='Cut from washable silk':fontsize=54:fontcolor=white:x=110:y=h-170:box=1:boxcolor=black@0.47:boxborderw=18" \ -c:v libx264 -crf 17 part-1.mp4 ffmpeg -y -ss 0 -t 3 -i shot-2.mp4 -an \ -vf "scale=1920:1080:flags=lanczos,fps=24,setsar=1,format=yuv420p,drawtext=fontfile=title.ttf:text='Soft enough to forget it’s there':fontsize=54:fontcolor=white:x=110:y=h-170:box=1:boxcolor=black@0.47:boxborderw=18" \ -c:v libx264 -crf 17 part-2.mp4 ffmpeg -y -ss 0 -t 3 -i shot-3.mp4 -an \ -vf "scale=1920:1080:flags=lanczos,fps=24,setsar=1,format=yuv420p,drawtext=fontfile=title.ttf:text='From studio to sofa':fontsize=54:fontcolor=white:x=110:y=h-170:box=1:boxcolor=black@0.47:boxborderw=18" \ -c:v libx264 -crf 17 part-3.mp4 # 2. End card: the product still with a slow push-in, a soft light fade on the left and the brand. ffmpeg -y -loop 1 -i product.png -f lavfi -i "color=c=0xFAF6F0:s=1920x1080:r=24" -filter_complex \ "[0:v]scale=3840:2160,zoompan=z='1+0.0012*on':x='iw/2-(iw/zoom/2)':y='ih/2-(ih/zoom/2)':d=84:s=1920x1080:fps=24[p];[1:v]format=rgba,geq=r='r(X,Y)':g='g(X,Y)':b='b(X,Y)':a='if(lt(X,1190),215*pow(1-X/1190,1.4),0)'[s];[p][s]overlay=0:0:shortest=1,drawtext=fontfile=title.ttf:text='SOFT HOURS':fontsize=94:fontcolor=0x281e20:x=130:y=366:alpha='min(1,max(0,(t-0.3)/0.4))',drawtext=fontfile=text.ttf:text='Silk Lounge Set':fontsize=52:fontcolor=0x5f5052:x=136:y=490:alpha='min(1,max(0,(t-0.3)/0.4))',drawtext=fontfile=text.ttf:text='Made for slow mornings.':fontsize=40:fontcolor=0xb05c68:x=136:y=575:alpha='min(1,max(0,(t-0.3)/0.4))',format=yuv420p" \ -t 3.5 -c:v libx264 -crf 17 end-card.mp4 # 3. Join the parts with hard cuts. printf "file 'part-1.mp4'\nfile 'part-2.mp4'\nfile 'part-3.mp4'\nfile 'end-card.mp4'\n" > parts.txt ffmpeg -y -f concat -safe 0 -i parts.txt -c copy ad-silent.mp4 # 4. Lay the music under it: fade in 0.5 s, fade out over the last 1.5 s. ffmpeg -y -i ad-silent.mp4 -i music.mp3 -filter_complex \ "[1:a]atrim=0:12.5,asetpts=PTS-STARTPTS,afade=t=in:st=0:d=0.5,afade=t=out:st=11:d=1.5,volume=0.9[a]" \ -map 0:v -map "[a]" -c:v copy -c:a aac -b:a 192k -t 12.5 -movflags +faststart wu-sefei-ad.mp4
Make your own ad with Wu Sefei
Follow the steps above with your own product: swap the product still and the shot ideas, keep the character description and the character reference.
More examples
All examplesPITWAVE · Team Radio Earpiece
With Liu Xialan · 4 shots · 16s
LITTLE HEARTH · Strawberry Cream Puffs
With Wen Li · 3 shots · 13s
GREENLINE · Whey Protein + Shaker
With Han Daozhen · 3 shots · 12s