FLF2V: First-Last-Frame-to-Video Generation with Wan2.1
FLF2V by GeradeHouse generates looping videos from start and end frames using Wan2.1. Users provide first and last images; the model synthesizes intermediate motion at 720p.
FLF2V
Wan2.1Video GenerationFirst-Last-FrameImage-to-VideoLoopingFirst-Last-Frame-to-Video generation by GeradeHouse, built on Wan2.1. Generates smooth looping videos by synthesizing intermediate frames between user-provided start and end images at 720p resolution.
| Developer | GeradeHouse |
| Architecture | Wan2.1-based |
| Model Sizes | 14B (fp16/fp8) |
| Capabilities | First-last-frame video interpolation, looping videos |
Overview
FLF2V (First-Last-Frame-to-Video) by GeradeHouse enables a unique video generation workflow where users provide only two images: the starting frame and the ending frame. The Wan2.1-based model then automatically synthesizes all intermediate frames, creating a smooth, logically coherent transition video at 720p resolution. This approach is ideal for creating looping videos, seamless transitions, and controlled motion sequences.
The model accepts both images along with an optional text prompt to guide the style and content of the interpolated motion.
ComfyUI Integration
FLF2V is supported in ComfyUI via the Kijai WanVideoWrapper. The fp8 variant provides efficient inference on consumer GPUs while maintaining high-quality output.
Available Sizes
| Size | Precision | Description |
|---|---|---|
| 14B | fp16 | Full precision FLF2V generation |
| 14B | fp8 | Quantized for efficient inference |
Resources
| Resource | Link |
|---|---|
| Hugging Face Space | GeradeHouse/Wan2.1-FLF2V |
| ComfyUI Tutorial | docs.comfy.org/tutorials/video/wan/wan-flf |
| Hugging Face (Kijai) | Kijai/WanVideo_comfy |
Guides and workflows related to this model series.