floyo logo
Powered by
ThinkDiffusion
⚡️Nano Banana 2 ⚡️ just landed. Start creating now.
floyo logo
Powered by
ThinkDiffusion
⚡️Nano Banana 2 ⚡️ just landed. Start creating now.

Vertical Video Prop & Object Replacement Using Seedream + Wan 2.2

131

Generates in about 13 secs

Nodes & Models

Seedream4Unified_floyo
INTConstant
INTConstant
ImageConcatMulti
ImageResizeKJv2
GetImageSizeAndCount
GrowMaskWithBlur
BlockifyMask
DrawMaskOnImage
DownloadAndLoadSAM2Model
sam2.1_hiera_base_plus.safetensors
Sam2Segmentation
DownloadAndLoadSAM2Model
sam2.1_hiera_base_plus.safetensors
Sam2Segmentation
OnnxDetectionModelLoader
PoseAndFaceDetection
DrawViTPose
GetNode
CLIPVisionLoader
clip_vision_h.safetensors
LoadImage
Note
Label (rgthree)
Fast Groups Bypasser (rgthree)
SetNode
SaveImage
PreviewImage
WanVideoContextOptions
WanVideoVAELoader
Wan2_1_VAE_bf16.safetensors
WanVideoLoraSelectMulti
WanAnimate_relight_lora_fp16.safetensors
lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
WanVideoBlockSwap
WanVideoTorchCompileSettings
WanVideoTextEncodeCached
umt5-xxl-enc-bf16.safetensors
WanVideoClipVisionEncode
WanVideoModelLoader
Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors
WanVideoAnimateEmbeds
WanVideoSetLoRAs
WanVideoSetBlockSwap
WanVideoSampler
WanVideoDecode
WanVideoVAELoader
Wan2_1_VAE_bf16.safetensors
WanVideoLoraSelectMulti
WanAnimate_relight_lora_fp16.safetensors
lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
WanVideoBlockSwap
WanVideoTorchCompileSettings
WanVideoModelLoader
Wan2_2-Animate-14B_fp8_e4m3fn_scaled_KJ.safetensors
WanVideoSetLoRAs
VHS_LoadVideo
VHS_VideoCombine
ImageConcatMulti
ImageResizeKJv2
GetImageSizeAndCount
DrawMaskOnImage

Performance-Driven Prop Replacement

This workflow lets you replace any prop in a video—like a watch, bottle, phone, or product—while keeping the original performance exactly the same.

You don’t need to reshoot anything.
The system updates the object, but preserves every detail of the actor’s movement:

✔ Same face and expressions
✔ Same hand motion
✔ Same timing
✔ Same body performance

Only the prop changes—cleanly and consistently.


How It Works

1. Turn on Seedream (the Image Editor)

Seedream acts like Photoshop for a single frame.
It lets you modify the first frame of the video, replacing the original object with the new one while preserving the scene’s realism.


2. Select the Area That Needs Replacement

Load the first frame and highlight the region:

  • the wrist

  • the hand

  • the table

  • or wherever the prop is located

This tells the system:
“This is where the new item will appear.”


3. Describe the New Prop

Tell Seedream exactly what should replace the old object.

Examples:

  • “gold luxury watch on the wrist”

  • “sleek black smartphone in hand”

  • “matte black perfume bottle on the table”

Seedream generates a clean, realistic edit with proper:
✔ angle
✔ lighting
✔ shadows
✔ reflections
✔ proportions


4. Send This Edited Frame Into Wan 2.2

The edited frame becomes the reference image.
Wan 2.2 understands:
“This is the new object—keep it stable in every frame of the video.”


5. Upload Your Original Video

Your raw video supplies all the natural body motion:

✔ hand movement
✔ wrist rotation
✔ expressions
✔ timing
✔ gestures

Wan 2.2 uses this motion data so the performance stays 100% original.


6. Activate Wan 2.2 Animate

Wan 2.2 rebuilds the entire video with the new prop integrated frame-by-frame.

The result:

✔ No flicker
✔ No jitter
✔ No warping
✔ The prop stays locked to the hand/wrist
✔ Motion stays perfectly natural


🎬 Final Result

You get a seamless video where:

✨ The original prop is fully replaced
✨ The new object looks physically correct
✨ The movement, expressions, and performance remain untouched
✨ The replacement stays consistent across the entire clip

The final output looks like the video was shot with the new prop from the beginning.


Read more

N