SnJake Sapsan-VFI is a custom ComfyUI node for x2 video frame interpolation. It inserts one in-between frame for every input pair, doubling the FPS.

---

Examples

https://github.com/user-attachments/assets/3730d3f7-687e-47de-a6e4-353493264a8c

https://github.com/user-attachments/assets/2f591a99-2789-426d-b4a3-7844f66f1101

https://github.com/user-attachments/assets/1ae8d400-6758-44f6-bd56-fc17d3f3bc06

---

Installation

The installation consists of two steps: installing the node and making the weights available.

Step 1: Install the Node

1. Open a terminal or command prompt.
2. Navigate to your ComfyUI custom_nodes directory.

   # Example for Windows
   cd D:\ComfyUI\custom_nodes\
   
   # Example for Linux
   cd ~/ComfyUI/custom_nodes/
   

3. Clone this repository:

   git clone https://github.com/SnJake/SnJake_Sapsan-VFI.git
   

4. For standard ComfyUI installations (with venv):
   1. Make sure your ComfyUI virtual environment (venv) is activated.
   2. Navigate into the new node directory and install the requirements (if needed):

      cd SnJake_Sapsan-VFI
      pip install -r requirements.txt
      

   For Portable ComfyUI installations:
   1. Navigate back to the root of your portable ComfyUI directory (e.g., D:\ComfyUI_windows_portable).
   2. Run the following command to use the embedded Python to install the requirements. Do not activate any venv.

      python_embeded\python.exe -m pip install -r ComfyUI\custom_nodes\SnJake_Sapsan-VFI\requirements.txt
      

Step 2: Model Weights

On first use the node can automatically download weights from the repository.

• Default weights location: ComfyUI/models/sapsan_vfi/
• Available weights on HF: Sapsan-VFI.safetensors and Sapsan-VFI.pt

If you want to download manually:

1. Download the weights from the HF repo: https://huggingface.co/SnJake/Sapsan-VFI.
2. Place the file(s) into ComfyUI/models/sapsan_vfi/.

Step 3: Restart

Restart ComfyUI completely. The node will appear under 😎 SnJake/VFI.

---

Usage

The node menu path is 😎 SnJake/VFI.

Inputs

• weights_name: Select weights from the dropdown (auto-download if missing).
• images: Input frame batch (IMAGE) from your video loader.
• fps (optional): Input FPS from GetVideoComponents (used to output x2 FPS).
• t: Interpolation time (default 0.5).
• pad_multiple: Padding multiple (0 = use config.yaml).
• match_brightness: Match predicted frame brightness to inputs (reduces flicker).
• amp: Precision (auto, bf16, fp16, none).
• device: Device (auto, cuda, cpu).
• torch_compile: Enable torch.compile (optional optimization).
• console_progress: Print progress to the ComfyUI console.

Outputs

• images: Interpolated frames (x2 length minus 1).
• fps: Output FPS (input FPS * 2, or 0 if not provided).

---

Training Details

• Created out of curiosity and personal interest.
• Total epochs: 11
• Dataset: 2700 videos
• Shards: 151 shards of 1000 shadrs in each. 151 000 triplets.

Training code is included in training_code/ for reference.

---

Disclaimer

This project was made purely for curiosity and personal interest. The code was written by GPT-5.2 Codex.

---

License

MIT.