FunGen

FunGen is a Python-based tool that uses AI to generate Funscript files from VR and 2D POV videos. It enables fully automated funscript creation for individual scenes or entire folders of videos.

Join the Discord community for discussions and support: Discord Community

Note: The necessary YOLO models will also be available via the Discord.

DISCLAIMER

This project is still at the early stages of development. It is not intended for commercial use. Please, do not use this project for any commercial purposes without prior consent from the author. It is for individual use only.

Prerequisites

Before using this project, ensure you have the following installed:

Git https://git-scm.com/downloads/ or 'winget install --id Git.Git -e --source winget' from a command prompt for Windows users as described below for easy install of Miniconda.
FFmpeg added to your PATH or specified under the settings menu (https://www.ffmpeg.org/download.html)
Miniconda (https://www.anaconda.com/docs/getting-started/miniconda/install)

Easy install of Miniconda for Windows users: Click Start, type "cmd", right click on Command Prompt, and select "Run as administrator." Enter "winget install -e --id Anaconda.Miniconda3" and press enter. Miniconda should then download and install.

Installation

Start a miniconda command prompt

After installing Miniconda look for a program called "Anaconda prompt (miniconda3)" in the start menu (on Windows) and open it

Create the necessary miniconda environment and activate it

conda create -n VRFunAIGen python=3.11
conda activate VRFunAIGen

Please note that any pip or python commands related to this project must be run from within the VRFunAIGen virtual environment.

Clone the repository

Open a command prompt and navigate to the folder where you'd like FunGen to be located. For example, if you want it in C:\FunGen, navigate to C:\ ('cd C:'). Then run

git clone --branch v0.5.0 https://github.com/ack00gar/FunGen-AI-Powered-Funscript-Generator.git FunGenBeta
cd FunGenBeta

If you have the original FunGen installed, skip to Download the YOLO model

NVIDIA GPU Setup (CUDA Required)

Quick Setup:

Install NVIDIA Drivers: Download here
Install CUDA 12.8: Download here
Install cuDNN for CUDA 12.8: Download here (requires free NVIDIA account)

Install Python Packages:

For 20xx, 30xx and 40xx-series NVIDIA GPUs:

uv sync --extra cuda

For 50xx series NVIDIA GPUs (RTX 5070, 5080, 5090):

uv sync --extra cuda-rtx50

Note: NVIDIA 10xx series GPUs are not supported.

Verify Installation:

nvidia-smi                    # Check GPU and driver
nvcc --version               # Check CUDA version
python -c "import torch; print(torch.cuda.is_available())"  # Check PyTorch CUDA

If your GPU doesn't support cuda

uv sync --extra cpu

If your GPU supports ROCm (AMD Linux Only)

uv sync --extra rocm

Download the YOLO model

Go to our discord to download the latest YOLO model for free. When downloaded place the YOLO model file(s) in the models/ sub-directory. If you aren't sure you can add all the models and let the app decide the best option for you.

Download the pose model

Download from https://docs.ultralytics.com/tasks/pose/ and place in the models/ sub-directory.

Start the app

uv run main.py

We support multiple model formats across Windows, macOS, and Linux.

Recommendations

NVIDIA Cards: we recommend the .engine model
AMD Cards: we recommend .pt (requires ROCm see below)
Mac: we recommend .mlmodel

Models

.pt (PyTorch): Requires CUDA (for NVIDIA GPUs) or ROCm (for AMD GPUs) for acceleration.
.onnx (ONNX Runtime): Best for CPU users as it offers broad compatibility and efficiency.
.engine (TensorRT): For NVIDIA GPUs: Provides very significant efficiency improvements (this file needs to be build by running "Generate TensorRT.bat" after adding the base ".pt" model to the models directory)
.mlpackage (Core ML): Optimized for macOS users. Runs efficiently on Apple devices with Core ML.

In most cases, the app will automatically detect the best model from your models directory at launch, but if the right model wasn't present at this time or the right dependencies where not installed, you might need to override it under settings. The same applies when we release a new version of the model.

AMD GPU acceleration

Coming soon

Troubleshooting CUDA Installation

Common Issues:

Driver version mismatch: Ensure NVIDIA drivers are compatible with your CUDA version
PATH issues: Make sure CUDA bin directory is in your system PATH
Version conflicts: Ensure all components (driver, CUDA, cuDNN, PyTorch) are compatible versions

Verification Commands:

nvidia-smi                    # Check GPU and driver
nvcc --version               # Check CUDA version  
python -c "import torch; print(torch.cuda.is_available())"  # Check PyTorch CUDA
python -c "import torch; print(torch.backends.cudnn.is_available())"  # Check cuDNN

GUI Settings

Find the settings menu in the app to configure optional option.

Start script

You can use Start windows.bat to launch the gui on windows.

GitHub Token Setup (Optional)

FunGen includes an update system that allows you to download and switch between different versions of the application. To use this feature, you'll need to set up a GitHub Personal Access Token.

Why a GitHub Token?

GitHub's API has rate limits:

Without a token: 60 requests per hour
With a token: 5,000 requests per hour

This allows FunGen to fetch commit information, changelogs, and version data without hitting rate limits.

How to Get a GitHub Token

Go to GitHub Settings:
- Visit GitHub Settings
- Sign in to your GitHub account
Navigate to Developer Settings:
- Click your GitHub avatar (top right) → "Settings"
- Scroll down to the bottom left of the Settings page
- Click "Developer settings" in the left menu list
Create a Personal Access Token:
- Click "Personal access tokens" → "Tokens (classic)"
- Click "Generate new token" → "Generate new token (classic)"
Confirm Access
- If you created a 2FA you will be prompted to eter it
- If you have not yet created a 2FA you will be prompted to do so
Configure the Token:
- Note: Give it a descriptive name like "FunGen Updates"
- Expiration: Choose an appropriate expiration (30 days, 60 days, etc.)
- Scopes: Select only these scopes:
  - public_repo (to read public repository information)
  - read:user (to read your user information for validation)
Generate and Copy:
- Click "Generate token"
- Important: Copy the token immediately - you won't be able to see it again!

Setting the Token in FunGen

Open FunGen and go to the Updates menu
Click "Select Update Commit"
Go to the "GitHub Token" tab
Paste your token in the text field
Click "Test Token" to verify it works
Click "Save Token" to store it

What the Token is Used For

The GitHub token enables these features in FunGen:

Version Selection: Browse and download specific commits from the v0.5.0 branch
Changelog Display: View detailed changes between versions
Update Notifications: Check for new versions and updates
Rate Limit Management: Avoid hitting GitHub's API rate limits

Security Notes

The token is stored locally in github_token.ini
Only public_repo and read:user permissions are required
The token is used only for reading public repository data
You can revoke the token anytime from your GitHub settings

Command Line Usage

FunGen can be run in two modes: a graphical user interface (GUI) or a command-line interface (CLI) for automation and batch processing.

To start the GUI, simply run the script without any arguments:

python main.py

To use the CLI mode, you must provide an input path to a video or a folder.

CLI Examples

To generate a script for a single video with default settings (3-stage mode):

python main.py "/path/to/your/video.mp4"

To process an entire folder of videos recursively using 2-stage mode and overwrite existing funscripts:

python main.py "/path/to/your/folder" --mode 2-stage --overwrite --recursive

Command-Line Arguments

Argument	Short	Description
`input_path`		Required for CLI mode. Path to a single video file or a folder containing videos.
`--mode`		Sets the processing mode. Choices: `2-stage`, `3-stage`, `oscillation-detector`. Default is `3-stage`.
`--overwrite`		Forces the app to re-process and overwrite any existing funscripts. By default, it skips videos that already have a funscript.
`--no-autotune`		Disables the automatic application of Ultimate Autotune after generation.
`--no-copy`		Prevents saving a copy of the final funscript next to the video file. It will only be saved in the application's output folder.
`--recursive`	`-r`	If the input path is a folder, this flag enables scanning for videos in all its subdirectories.

Performance & Parallel Processing

Our pipeline's current bottleneck lies in the Python code within YOLO.track (the object detection library we use), which is challenging to parallelize effectively in a single process.

However, when you have high-performance hardware you can use the command line (see above) to processes multiple videos simultaneously. Alternatively you can launch multiple instances of the GUI.

We tested speeds of about 60 to 110 fps for 8k 8bit vr videos when running a single process. Which translates to faster then realtime processing already. However, running in parallel mode we tested speeds of about 160 to 190 frames per second (for object detection). Meaning processing times of about 20 to 30 minutes for 8bit 8k VR videos for the complete process. More then twice the speed of realtime!

Keep in mind your results may vary as this is very dependent on your hardware. Cuda capable cards will have an advantage here. However, since the pipeline is largely CPU and video decode bottlenecked a top of the line card like the 4090 is not required to get similar results. Having enough VRAM to run 3-6 processes, paired with a good CPU, will speed things up considerably though.

Important considerations:

Each instance requires the YOLO model to load which means you'll need to keep checks on your VRAM to see how many you can load.
The optimal number of instances depends on a combination of factors, including your CPU, GPU, RAM, and system configuration. So experiment with different setups to find the ideal configuration for your hardware! 😊

Miscellaneous

For VR only sbs (side by side) Fisheye and Equirectangular 180° videos are supported at the moment
2D POV videos are supported but work best when they are centered properly
2D / VR is automatically detected as is fisheye / equirectangular and FOV (make sure you keep the file format information in the filename _FISHEYE190, _MKX200, _LR_180, etc.)
Detection settings can also be overwritten in the UI if the app doesn't detect it properly

Output Files

The script generates the following files in a dedicated subfolder within your specified output directory:

_preprocessed.mkv: A standardized video file used by the analysis stages for reliable frame processing.
.msgpack: Raw YOLO detection data from Stage 1. Can be re-used to accelerate subsequent runs.
_stage2_overlay.msgpack: Detailed tracking and segmentation data from Stage 2, used for debugging and visualization.
_t1_raw.funscript: The raw, unprocessed funscript generated by the analysis before any enhancements are applied.
.funscript: The final, post-processed funscript file for the primary (up/down) axis.
.roll.funscript: The final funscript file for the secondary (roll/twist) axis, generated in 3-stage mode.
.fgp (FunGen Project): A project file containing all settings, chapter data, and paths related to the video.

About the project

Pipeline Overview

The pipeline for generating Funscript files is as follows:

YOLO Object Detection: A YOLO model detects relevant objects (e.g., penis, hands, mouth, etc.) in each frame of the video.
Tracking and Segmentation: A custom tracking algorithm processes the YOLO detections to identify and segment continuous actions and interactions over time.
Funscript Generation: Based on the mode (2-stage, 3-stage, etc.), the tracked data is used to generate a raw Funscript file.
Post-Processing: The raw Funscript is enhanced with features like Ultimate Autotune to smooth motion, normalize intensity, and improve the overall quality of the final .funscript file.

Project Genesis and Evolution

This project started as a dream to automate Funscript generation for VR videos. Here’s a brief history of its development:

Initial Approach (OpenCV Trackers): The first version relied on OpenCV trackers to detect and track objects in the video. While functional, the approach was slow (8–20 FPS) and struggled with occlusions and complex scenes.
Transition to YOLO: To improve accuracy and speed, the project shifted to using YOLO object detection. A custom YOLO model was trained on a dataset of 1000nds annotated VR video frames, significantly improving detection quality.
Original Post: For more details and discussions, check out the original post on EroScripts: VR Funscript Generation Helper (Python + CV/AI)

Contributing

Contributions are welcome! If you'd like to contribute, please follow these steps:

Fork the repository.
Create a new branch for your feature or bug fix.
Commit your changes.
Submit a pull request.

License

This project is licensed under the Non-Commercial License. You are free to use the software for personal, non-commercial purposes only. Commercial use, redistribution, or modification for commercial purposes is strictly prohibited without explicit permission from the copyright holder.

This project is not intended for commercial use, nor for generating and distributing in a commercial environment.

For commercial use, please contact me.

See the LICENSE file for full details.

Acknowledgments

YOLO: Thanks to the Ultralytics team for the YOLO implementation.
FFmpeg: For video processing capabilities.
Eroscripts Community: For the inspiration and use cases.

Support

If you encounter any issues or have questions, please open an issue on GitHub.