Turn one full-body photo into a week of outfits and an outfit-swap video

What you'll get

Start with one full-body photo and make five work outfits for the same person, one for each weekday. Then turn one of those looks into a short video where the person changes clothes in place. You can post the video on TikTok, Reels, or similar apps. You can also use the images as model photos for an online store. Everything is done on the Dreamina website by clicking through the steps. No coding needed. This post includes 5 prompts you can copy and paste.

The AI model makes the images and the video. You pick the reference photos and check that the face still looks like the real person. If the face drifts, every step after it drifts too, so spend the most time on step 1.

Before you start: use only your own photos

This method uses a real person's face, so keep these rules in mind:

• Only use your own photos, or photos of someone who gave written permission. Don't use photos of celebrities, influencers, or friends who didn't agree. Posting a swapped version could violate their right to their own image, and the platform may remove it.

• Label the video as AI-generated when you post it. TikTok, Instagram, and other platforms have AI content labels. Adding the label yourself works better than having the system detect it and limit your reach.

• You can copy clothing ideas from others, but don't copy brand logos. Step 2 pulls the clothes out of an outfit photo. If you plan to use the result commercially, remove any obvious brand marks or replace them with your own.

Step 1: Build a base portrait from two photos

Open the image generator in Dreamina and pick the newest Seedream model. This guide uses Seedream 5.5. Upload two photos together:

• A close-up photo. Your facial features should be clear. This locks in the face.

• A full-body photo. This locks in the body shape and standing pose.

With only a full-body photo, the face is too small in the frame. There isn't enough detail to work with, and after a few tries the person starts to look different. Two photos give each part its own source: one for the face, one for the body.

Put the close-up as Image 1 and the full-body photo as Image 2. Then send this prompt:

Image 1 is the face reference for the person. Image 2 is the reference for body shape and standing pose.
Generate a full-body photo of the same person:
1. Keep the face, facial features, and hairstyle exactly as in Image 1. Don't beautify or change the face shape.
2. Follow the body proportions and standing pose from Image 2.
3. Clothing: [light silk pajamas]. Setting: [bedroom by the window, natural daylight].
4. Vertical 9:16 format, front-facing full body, with a natural shadow under the feet.

This image is the base for every outfit swap that follows. The pajamas are just an example. You can use any look you want as the "before" outfit. When the result comes out, zoom in on the face. If it doesn't look like the real person, generate it again. Don't settle for "close enough."

Step 2: Split the reference outfits into white-background item images

If you don't know how to put outfits together, that's fine. Find an outfit reference instead. You can use online store model photos, street style photos, or pictures of clothes you already own. Give the image to the model and ask it to keep only the clothes, shoes, and bag. It will turn them into a white-background "ghost mannequin" image, where the clothes look like they're worn by an invisible person, so you can still see the cut.

Extract the clothing worn by the person in this image as separate items.
1. Include the top, bottoms, shoes, and bag. Remove the person and show the items as a ghost mannequin product image on a white background.
2. Keep the color, fabric texture, fit, and length of each item unchanged.
3. Remove any brand logos and text from the clothes.
4. Front-facing view, vertical composition.

You need five outfits for five days. Once the items are split out, the model only has to look at the clothes during the swap. It won't be pulled off track by the face or pose of the person in the original photo.

Step 3: Swap one outfit at a time, always starting from the base

This step is the core of the whole process. Upload two images. Image 1 is the base portrait. Image 2 is the item image for one day.

A few notes on why the prompt is written this way:

• The line "Don't copy the pose from Image 2" is there on purpose. The mannequin image has its own pose. If you don't say so, the model often poses the person like the mannequin.

• Tuesday through Friday still start from the same base portrait. Don't use Monday's result as the starting point for Tuesday. Each edit on top of the last one adds a bit more error, and by Friday the face may have changed.

• Keep the prompt word for word and only swap Image 2. That keeps the lighting and camera angle the same across all five images, so they look like one photo shoot.

When all five images are ready, zoom in on each face and compare them side by side. If one doesn't look right, redo just that one.

Step 4: Generate an empty room with no people

The video needs three images: the outfit before, the outfit after, and an empty room. You already have the first two. For the third, edit the base portrait:

Remove the person from this image and keep only the empty room.
Keep the room layout, furniture, windows, light, and shadow direction the same. Fill in the floor and background where the person was standing, so it looks natural.

The empty room gives the video model a reference that says "the setting must not change." That way, the walls and furniture are less likely to wobble during the outfit change.

Step 5: Turn the outfit swaps into one video

In Dreamina, switch to video generation and pick Seedance. This guide uses Seedance 2.5. Upload all three images as references, then describe what each one does:

Image 1 is the starting outfit. Image 2 is the outfit after the change. Image 3 is the scene reference.
Generate a [5]-second vertical video. The scene always matches Image 3, and the camera stays still:
1. At the start, the person wears the outfit from Image 1, standing in place and looking naturally at the camera.
2. Outfit change: [the person turns around in place, and by the time they face the camera again, they are wearing the outfit from Image 2].
3. The video ends on the look from Image 2, with the pose matching Image 2.
The person's face stays consistent the whole time, with no distortion.

How you hide the outfit change decides whether the video looks natural. The most common method is a full turn in place. You can also cover the camera with a piece of clothing and then move it away, or have the person walk out of frame and back in. Another option is a fast camera whip that snaps away and back. If your account has a "first and last frame" mode, you can set Image 1 as the first frame and Image 2 as the last frame.

Want a more detailed prompt, such as the rhythm of the movement or the expression? Paste the prompt above along with your three images into any AI assistant, and ask it to expand the prompt for your outfit.

Editing many rounds without drifting off course

When you sell clothes online, you often edit the same piece several times. You might change the color, then the fabric, and then add text. This checklist can help.

• Change only one thing per round. Add this line at the end of the prompt: "Change only this one thing. Don't change anything else."

• Every three or four rounds, put the image next to the original and compare the faces. If the face has changed, go back to the last image that still looked right and edit from there.

• To add text to clothing, put the words in quotation marks and say where they go. For example: Add the text "YOUR BRAND HERE" in white, in a small size, at the center of the sweater's chest. When you edit something else later, add this line: "Keep the text on the chest."

• If you need assets for a poster or product page, ask the model to extract the bag or shoes as transparent PNGs. That way you won't have to cut them out by hand when you build the layout.

Notes

• Models roll out at different times on different accounts. If your Dreamina doesn't show the versions named in this guide, use the newest version available. The steps are the same.

• Image and video generation both use credits, and video costs the most. Lock in your five outfit images first, then make the video, so you waste fewer tries.

• The face is the easiest part to go wrong. Zoom in and check it at every step. You're using real people's photos, so confirm you have permission and add the AI label before you publish.

• Small text on clothing sometimes comes out with typos. Check every letter before you use the image commercially.

References

• Official Dreamina guide: How to Access Seedance 2.5 in Dreamina

• Official Dreamina guide: Seedance 2.5 Prompts Guide

• Official Dreamina guide: First Last Frame

Want to pick a pose for your character first? See Make AI video characters stop moving like puppets with a one-tap Muse pose library.