pipeline-examples/flux-to-wan

Pass FLUX-finetuned image outputs into WAN 2.1

Public
3 runs

Readme

flux-to-wan

Generate videos from descriptions using Flux-finetuned image models and WAN 2.1 video generation.

https://replicate.com/pipelines-beta/flux-finetune-image-to-wan-video

See also:

  • flux-lora-video – basically the same thing, but with Claude-powered prompt enhancement, and it uses Minimax’s video generation model instead of WAN 2.1

Features

  • Create high-quality videos from text prompts
  • Use any Flux-finetuned model as the image source
  • Seamlessly converts still images into smooth 720p video

Models

Under the hood it uses these models:

  • wavespeedai/wan-2.1-i2v-720p: A powerful image-to-video model that transforms still images into cinematic videos
  • Flux finetuned models (default: zeke/ziki-flux): Custom image generation models trained using Flux

How it works

This pipeline takes your text prompt and passes it to a Flux-finetuned image model to generate a still image. This image is then fed into WAN 2.1, along with your original prompt, to create a fluid video animation that brings your concept to life. Just remember to include your Flux model’s trigger word in the prompt for best results.