Example Windows 11 Install Steps April 20, 2024 - jhogsett/EMA-VFI-WebUI GitHub Wiki

Pre-requisites

install git for windows (to get code from the GitHub repos) https://git-scm.com/download/win
install anaconda for windows (to run Python) https://www.anaconda.com/download

Set Up Anaconda Environment

open an anaconda console window (via Windows start menu)
- Anaconda3 -> Anaconda Prompt (anaconda3)
create a new directory for the app, and change into it

(base) C:\>
(base) C:\>md conda
(base) C:\>cd conda
(base) C:\conda>md EMA-VFI-WebUI
(base) C:\conda>cd EMA-VFI-WebUI

create a new anaconda environment
- see https://conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#creating-an-environment-with-commands

(base) C:\conda\EMA-VFI-WebUI>conda create --name EMA-VFI-WebUI python=3.10.9

activate the new environment
- see https://conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#activating-an-environment

(base) C:\conda\EMA-VFI-WebUI>conda activate EMA-VFI-WebUI

Set Up the EMA-VFI Engine

clone the EMA-VFI engine

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>git clone https://github.com/MCG-NJU/EMA-VFI.git

change into the newly cloned directory

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>cd EMA-VFI

create a directory for the model checkpoints

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>md ckpt

download the model checkpoints
- Go to Google Drive at https://drive.google.com/drive/folders/16jUa3HkQ85Z5lb5gce1yoaWkP-rdCd0o?usp=sharing
- download these files and place into the C:\conda\EMA-VFI-WebUI\EMA-VFI\ckpt directory
  - ours.pkl (251.5MB)
  - ours_t.pkl (251.5MB)

Install NVIDIA Cuda if necessary

Determine if cuda is installed and find cuda version:
- version 12.4 or later is recommended

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Thu_Mar_28_02:30:10_Pacific_Daylight_Time_2024
Cuda compilation tools, release 12.4, V12.4.131
Build cuda_12.4.r12.4/compiler.34097967_0

If not installed, or version is < 12.4
- Go to https://developer.nvidia.com/cuda-downloads?target_os=Windows&target_arch=x86_64&target_version=11&target_type=exe_local
- Double-check the particular installer you need, for instance your Windows version may differ

Install EMA-VFI depenencies (updated somewhat from their recommended versions)

Install Torch 2.2.0

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>conda install pytorch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 pytorch-cuda=12.1 -c pytorch -c nvidia

Install scikit-image (latest 0.22.0)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>conda install scikit-image

Install numpy (latest 1.26.4)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>conda install numpy

Install opencv-python (latest 4.9.0.80)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>pip install opencv-python

Install timm (latest 0.9.16)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>pip install timm

Install tqdm (latest 4.66.2)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>conda install tqdm

Install imageio (downgrade to 2.24.0 due to use of deprecated kwargs)

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>pip install imageio==2.24.0

Check the EMA-VFI Engine installation by running an inference:

Run the basic (non-time-based) demo:

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI>python demo_2x.py

If it worked, there will be a new file out_2x.gif in this directory:
- C:\conda\EMA-VFI-WebUI\EMA-VFI\example
Open it to confirm it created an animated gif of a toy train

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI\example>start out_2x.gif

Install Real-ESRGAN

Go back to the application root directory

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI\example>cd c:\conda\EMA-VFI-WebUI

Clone the repo

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>git clone https://github.com/xinntao/Real-ESRGAN.git

change into the newly cloned directory

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>cd Real-ESRGAN

install the dependencies

(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\Real-ESRGAN>pip install basicsr
(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\Real-ESRGAN>pip install facexlib
(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\Real-ESRGAN>pip install gfpgan
(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\Real-ESRGAN>pip install -r requirements.txt
(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\Real-ESRGAN>python setup.py develop

IMPORTANT You may need to patch basicsr due to a changed dependent module name

If you see this error when running the application:

ModuleNotFoundError: No module named 'torchvision.transforms.functional_tensor'

Do the following:

Edit the file degradations.py, located for example in C:\Users\USER\.conda\envs\avr\lib\site-packages\basicsr\data\degradations.py
Change the line from torchvision.transforms.functional_tensor import rgb_to_grayscale to from torchvision.transforms.functional import rgb_to_grayscale (See: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/13985)

NOTE: Real-ESRGAN will download the 4x+ model on first use

Set Up EMA-VFI WebUI

Go back to the application root directory

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI\EMA-VFI\example>cd c:\conda\EMA-VFI-WebUI

clone the EMA-VFI WebUI application

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>git clone https://github.com/jhogsett/EMA-VFI-WebUI.git

change into the newly cloned directory

(EMA-VFI-WebUI) C:\conda\EMA-VFI-WebUI>cd EMA-VFI-WebUI

install the dependencies

(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI>pip install -r requirements.txt

Merge EMA-VFI and Real-ESRGAN components into WebUI Application

Copy the following directories and files from: C:\conda\EMA-VFI-WebUI\EMA-VFI to: C:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI
- directories: benchmark, ckpt, model
- files: config.py, dataset.py, Trainer.py
Copy the following directories and files from: C:\conda\EMA-VFI-WebUI\Real-ESRGAN to: C:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI
- directory: realesrgan

Install FFmpeg

see https://ffmpeg.org/download.html#build-windows
I use the Windows builds at https://www.gyan.dev/ffmpeg/builds/

Start EMA-VFI WebUI Application

change to the application repo directory

cd C:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI

start the application

(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI>webui.bat

Testing an Interpolation

test files are location in c:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI\images
on the Interplolate Frames/Interpolate Frames tab
- drag image0.png from the images directory to the "before image" panel
- drag image2.png to the "after image" panel
click Interpolate
after a short delay, the Animated Preview panel should show a 3-frame animated GIF

Running EMA-VFI WebUI the next time

open an anaconda console window (via Windows start menu)
- Anaconda3 -> Anaconda Prompt (anaconda3)
activate the environment created before
- see https://conda.io/projects/conda/en/latest/user-guide/tasks/manage-environments.html#activating-an-environment

(base) C:\conda\EMA-VFI-WebUI>conda activate EMA-VFI-WebUI

change to the application repo directory

cd C:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI

start the application

(EMA-VFI-WebUI) c:\conda\EMA-VFI-WebUI\EMA-VFI-WebUI>webui.bat