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 Scene Extension & Coverage Generator using Seedream +Wan

208

Generates in about 32 secs

Nodes & Models

Seedream4Unified_floyo
Seed (rgthree)
ImageResizeKJv2
GetImageSizeAndCount
WanVideoVAELoader
Wan2_1_VAE_bf16.safetensors
WanVideoTorchCompileSettings
WanVideoLoraSelect
lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
WanVideoBlockSwap
LoadWanVideoT5TextEncoder
umt5-xxl-enc-bf16.safetensors
WanVideoModelLoader
wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors
wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors
WanVideoSetLoRAs
WanVideoVAELoader
Wan2_1_VAE_bf16.safetensors
WanVideoTorchCompileSettings
WanVideoLoraSelect
lightx2v_I2V_14B_480p_cfg_step_distill_rank64_bf16.safetensors
WanVideoBlockSwap
LoadWanVideoT5TextEncoder
umt5-xxl-enc-bf16.safetensors
WanVideoModelLoader
wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors
wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors
WanVideoTextEncode
CreateCFGScheduleFloatList
WanVideoSetBlockSwap
WanVideoSetLoRAs
WanVideoImageToVideoEncode
WanVideoSampler
WanVideoDecode
INTConstant
INTConstant
ImageResizeKJv2
GetImageSizeAndCount
Label (rgthree)
Note
EmptyLatentImage
LoadImage
Fast Groups Bypasser (rgthree)
PreviewImage
SaveImage
RIFE VFI
rife47.pth
VHS_VideoCombine

Scene Extension Workflow – Seedream + Wan

This workflow lets you extend a scene in a video—show more of the location, add new background details, or subtly move the camera—without reshooting.

The person, expressions, and motion stay the same. Only the background or environment is expanded.


Step 1: Generate the Extended Frame in Seedream

  1. Enable Image-to-Image mode in Seedream.

  2. Upload your first frame of the video.

  3. Enter a scene-extension prompt (for example: “extend terrace to the left, add Japanese garden, keep mansion in background, photorealistic, cinematic, ultra-detailed”).

  4. Seedream outputs an extended frame, which serves as the guide image for the new scene.


Step 2: Prepare Frames for Wan

  1. Use Frames-to-Video (Wan 2.2).

  2. Provide two images:

    • First frame → preserves the person, expressions, and gestures.

    • Extended frame → shows the new background / extended scene.

  3. Wan analyzes both frames to generate a smooth video transition.


Step 3: Generate the Extended Video

  1. Activate Wan and any fast processing options.

  2. Wan reconstructs the video with:

    • Person → unchanged

    • Expressions & motion → preserved

    • Camera movement → maintained

    • Background / scene → extended seamlessly

✅ The result is a clean, cinematic video where the scene is extended, the background looks richer, and the person’s motion is completely natural.


Tips for Best Results

  • Use high-resolution input frames in Seedream.

  • Keep consistent lighting between first and extended frames.

  • For smooth transitions, make your extension prompt very specific about perspective, camera pan, and elements like terraces, gardens, or architecture.

Read more

N