Innovate Futures @ Benji

Wan 2.1 First Last Frame To Video - Better Than Wan Fun Inp With 720P!

Added 2025-04-24 14:14:31 +0000 UTC

Video Tutorial : https://youtu.be/7w3QR5sEhZw

WAN 2.1's First & Last Frames To Video model—a revolutionary AI tool that transforms just two images into seamless video clips! In this video, we test its 14B-parameter, 720p-capable model in ComfyUI, comparing FP8 and FP16 versions for performance, and reveal how it outperforms older interpolation methods. Watch as we generate smooth transitions between uncorrelated images, explore tile upscaling tricks, and share pro tips for optimizing VRAM usage.

Who Is This For?

AI video creators wanting dynamic transitions from minimal inputs.

ComfyUI users exploring native WAN 2.1 integrations.

Tech enthusiasts curious about 14B-parameter AI models.

Filmmakers experimenting with AI-driven storyboarding.

Why It Matters:

This tool bridges the gap between static images and fluid video—ideal for prototyping scenes, ads, or creative projects. While still experimental, its ability to predict "in-between" frames opens new possibilities for rapid content creation.

Get Your Comfy Ready Before Get Started:

1- update your COMFYUI

git pull (for manual install)

or

update comfyui.bat (in Windows Portable version , I forgot the name of the Bat file , I haven't use Portable version long time)

2- Offical Wan2.1-FLF2V Model Weight, take a look of the concept before you get start : https://huggingface.co/Wan-AI/Wan2.1-FLF2V-14B-720P

3- Download Model file

For VRam 24GB or below : GGUF: https://huggingface.co/city96/Wan2.1-FLF2V-14B-720P-gguf/tree/main

From Comfy Org -

FP16

https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp16.safetensors

FP8

https://huggingface.co/Comfy-Org/Wan_2.1_ComfyUI_repackaged/blob/main/split_files/diffusion_models/wan2.1_flf2v_720p_14B_fp8_e4m3fn_scaled.safetensors

4 - Run It!

Attached ComfyUI Workflow file that use in this tutorial: