Qwenfluxprompt

🚀 Qwenfluxprompt

This is a LoRA for video generation, which can be used with the Wan2.1 14b video generation model. It offers a convenient way to generate videos based on text or image inputs.

✨ Features

• Compatibility: This LoRA can be used with diffusers or ComfyUI, and is compatible with both the text - to - video and image - to - video Wan2.1 models.
• Training Platform: It was trained on Replicate using the AI toolkit at https://replicate.com/ostris/wan - lora - trainer/train.
• Trigger Mechanism: Use the COLTOK trigger word to initiate video generation.

📦 Installation

This section doesn't have explicit installation steps in the traditional sense. However, if you want to use it with specific tools, here are the relevant commands:

Using with Diffusers

pip install git+https://github.com/huggingface/diffusers.git

💻 Usage Examples

Basic Usage

Run this LoRA with an API using Replicate

import replicate

input = {
    "prompt": "COLTOK",
    "lora_url": "https://huggingface.co/mam33/qwenfluxprompt/resolve/main/wan2.1-14b-coltok-lora.safetensors"
}

output = replicate.run(
    "fofr/wan2.1-with-lora:f83b84064136a38415a3aff66c326f94c66859b8ad7a2cb432e2822774f07b08",
    model="14b",
    input=input
)
for index, item in enumerate(output):
    with open(f"output_{index}.mp4", "wb") as file:
        file.write(item.read())

Advanced Usage

Using with Diffusers

import torch
from diffusers.utils import export_to_video
from diffusers import AutoencoderKLWan, WanPipeline
from diffusers.schedulers.scheduling_unipc_multistep import UniPCMultistepScheduler

model_id = "Wan-AI/Wan2.1-T2V-14B-Diffusers"
vae = AutoencoderKLWan.from_pretrained(model_id, subfolder="vae", torch_dtype=torch.float32)
pipe = WanPipeline.from_pretrained(model_id, vae=vae, torch_dtype=torch.bfloat16)
flow_shift = 3.0  # 5.0 for 720P, 3.0 for 480P
pipe.scheduler = UniPCMultistepScheduler.from_config(pipe.scheduler.config, flow_shift=flow_shift)
pipe.to("cuda")

pipe.load_lora_weights("mam33/qwenfluxprompt")

pipe.enable_model_cpu_offload() #for low-vram environments

prompt = "COLTOK"
negative_prompt = "Bright tones, overexposed, static, blurred details, subtitles, style, works, paintings, images, static, overall gray, worst quality, low quality, JPEG compression residue, ugly, incomplete, extra fingers, poorly drawn hands, poorly drawn faces, deformed, disfigured, misshapen limbs, fused fingers, still picture, messy background, three legs, many people in the background, walking backwards"

output = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    height=480,
    width=832,
    num_frames=81,
    guidance_scale=5.0,
).frames[0]
export_to_video(output, "output.mp4", fps=16)

🔧 Technical Details

• Steps: 2000
• Learning rate: 0.0001
• LoRA rank: 32

📄 License

This project is licensed under the Apache - 2.0 license.

Additional Information

Trigger words

You should use COLTOK to trigger the video generation.

Replicate Collections

Replicate has a collection of Wan2.1 models that are optimised for speed and cost. They can also be used with this LoRA:

• https://replicate.com/collections/wan - video
• https://replicate.com/fofr/wan2.1 - with - lora

Contribute your own examples

You can use the community tab to add videos that show off what you’ve made with this LoRA.