Skip to content
forked from banodoco/Dough

Dough is a open source tool for steering AI animations with precision.

License

Notifications You must be signed in to change notification settings

infiloop2/Dough

 
 

Repository files navigation

Welcome to Dough v. 0.9.1 (beta)

⬇️ Scroll down for Setup Instructions - Currently available on Linux & Windows, hosted version coming soon.

Dough is a tool for crafting videos with AI. Our goal is to give you enough control over video generations that you can make beautiful creations of anything you imagine that feel uniquely your own.

To achieve this, we allow you to guide video generations with precision using a combination of images (via Steerable Motion) examples videos (via Motion Director).

Below is brief overview and some examples of outputs:

With Dough, you can makes guidance frames using Stable Diffusion XL, IP-Adapter, Fooocus Inpainting, and more:

You can then assemble these frames into shots that you can granularly edit:

And then animate these shots by defining parameters for each frame and selecting guidance videos via Motion LoRAs:

As an example, here's a video that's guided with just images on high strength:

While here's a more complex one, with low strength images driving it alongside a guidance video:

And here's a more complex example combining high strength guidance with a guidance video strongly influencing the motion:

We're obviously very biased think that it'll be possible to create extraordinarily beautiful things with this and we're excited to see what you make! Please share stuff you made in our Discord.

Setup Instructions

Setting up on Runpod (click to expand)
  1. We recommend setting up persistent storage for a quick setup and for your projects to persist. To get it going, click into “Storage”, select “New Network Volume”. 50GB should be more than enough to start.

  2. Select a machine - any should work, but we recommend a 4090.

  3. During setup, open the relevant ports for Dough like below:

  1. When you’ve launched the pod, click into Jupyter Notebook:
  1. Follow the “Setup for Linux” below and come back here when you’ve gone through them.

  2. Once you’re done that, grab the IP Address for your instance:

Then form put these into this form with a : between them like this:

{Public ID}:{External Pair Value}

In the above example, that would make it:

213.173.108.4:14810

Then go to this URL, and it should be running!

Important: remember to terminate the instance once you’re done - you can restart it by following the instructions from step 3 above.

Instructions for Linux:

Install the app

This commands sets up the app. Run this only the first time, after that you can simply start the app using the next command.

curl -sSL https://raw.githubusercontent.com/banodoco/Dough/green-head/scripts/linux_setup.sh | bash

Enter the folder

In terminal, run:

cd Dough

Run the app

you can run the app using

source ./dough-env/bin/activate && ./scripts/entrypoint.sh

Instructions for Windows:

Open Powershell in Administrator mode

Open the Start menu, type Windows PowerShell, right-click on Windows PowerShell, and then select Run as administrator.

Install the app

Install MS C++ Redistributable (if not already present) - https://aka.ms/vs/16/release/vc_redist.x64.exe

Navigate to Documents

Make sure you're in the documents folder by running the following command:

cd ~\Documents

Run the setup script

iwr -useb "https://raw.githubusercontent.com/banodoco/Dough/green-head/scripts/windows_setup.bat" -OutFile "script.bat"
Start-Process "cmd.exe" -ArgumentList "/c script.bat"

Enter the folder

In Powershell, run:

cd Dough

Run the app

. .\dough-env\Scripts\activate ; .\scripts\entrypoint.bat

Using personal ComfyUI instance

You can provide the path to your ComfyUI instance and Dough will use the models present there. Add the following at the bottom of the ".env" file

COMFY_MODELS_BASE_PATH="full_path_to_your_ComfyUI"

for e.g. in the case of Linux users this may look something like this

COMFY_MODELS_BASE_PATH="/home/ubuntu/my_apps/ComfyUI/"

It will also download the new models in your personal ComfyUI instance. Please note that Dough will maintain it's own copy of the nodes and packages to not cause any issues with the packages that might already be installed on your personal ComfyUI.

If you're having any issues, please share them in our Discord.

Troubleshooting

Common problems (click to expand)
Issue during installation
  • Make sure you are using python3.10
  • If you are on Windows, make sure permissions of the Dough folder are not restricted (try to grant full access to everyone)
  • Double-check that you are not inside any system-protected folders like system32
  • Install the app in admin mode. Open the powershell in the admin mode and run "Set-ExecutionPolicy RemoteSigned". Then follow the installation instructions given in the readme
  • If all of the above fail, try to run the following instructions one by one and report which one is throwing the error
    call dough-env\Scripts\activate.bat
    python.exe -m pip install --upgrade pip
    pip install -r requirements.txt
    pip install websocket
    pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
    pip install -r comfy_runner\requirements.txt
    pip install -r ComfyUI\requirements.txt
Unable to locate credentials Make a copy of ".env.sample" and rename it to ".env"
Issue during runtime
  • If a particular node inside Comfy is throwing an error then delete that node and restart the app
  • Make sure you are using python3.10 and the virtual environment is activated
  • Try doing "git pull origin main" to get the latest code
Generations are in progress for a long time
  • Check the terminal if any progress is being made (they can be very slow, especially in the case of upscaling)
  • Cancel the generations directly from the sidebar if they are stuck
  • If you don't see any logs in the terminal, make sure no other program is running at port 12345 on your machine as Dough uses that port
Some other error?

Drop in our Discord.

About

Dough is a open source tool for steering AI animations with precision.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 99.0%
  • Other 1.0%