ComfyUI VFX Nodes
by FugitiveExpert01
A growing collection of custom ComfyUI nodes built for VFX production pipelines — designed around the real demands of working with high-resolution imagery, video sequences, and AI-assisted visual effects workflows.
📌 Overview
This repo fills the gaps between off-the-shelf ComfyUI nodes and the specific needs of VFX work: large format images, temporal consistency across video frames, precise spatial control, and clean integration with diffusion-based upscaling and enhancement models.
Nodes are built with video batch support as a first-class concern — not an afterthought.
🧩 Nodes
🔲 TileSplit
Splits an image or video batch into an overlapping grid of tiles, ready to be passed individually to a model.
| Parameter | Type | Description |
|---|---|---|
image |
IMAGE | Input image or video batch |
tiles_x |
INT | Number of columns |
tiles_y |
INT | Number of rows |
overlap_percent |
FLOAT | Overlap between adjacent tiles |
Outputs: tiles (LIST) · debug_image · tile_calc
Features:
- Overlap calculated automatically and stored for seamless reconstruction
- Outputs tiles as
(F, H, W, C)tensors — compatible with video model K-samplers - Debug image showing tile boundaries and overlap regions
- Tile dimensions snapped to multiples of 8 for model compatibility
🔳 TileMerge
Reconstructs a full image or video sequence from processed tiles using Laplacian Pyramid blending for seamless, artifact-free joins.
| Parameter | Type | Description |
|---|---|---|
tiles |
IMAGE (LIST) | Processed tile sequences |
tile_calc |
TILE_CALC | Layout data from TileSplit |
Outputs: image
Features:
- Gaussian feathering masks at overlap regions
- Laplacian frequency blending for natural-looking seams
- Robust handling of tensor shapes returned by video models
- Weighted accumulation handles overlapping regions correctly
🎬 Video Upscale With Model
Memory-efficient upscaling of every frame in a video batch using a ComfyUI upscale model (ESRGAN, RealESRGAN, etc.).
| Parameter | Type | Description |
|---|---|---|
model_name |
upscale_models | Model from your ComfyUI upscale_models folder |
images |
IMAGE | Input video batch (F, H, W, C) |
upscale_method |
ENUM | Final resize filter: nearest-exact, bilinear, area, bicubic |
factor |
FLOAT | Output scale multiplier (e.g. 2.0 = double resolution) |
device_strategy |
ENUM | See table below |
batch_size |
INT | Frames processed per GPU pass |
Device strategies:
| Strategy | Description |
|---|---|
auto |
Detects free VRAM and picks the best strategy automatically |
keep_loaded |
Model stays on GPU — fastest, requires available VRAM |
load_unload_each_frame |
Model moves to GPU per batch then back to CPU — lower peak VRAM |
cpu_only |
Runs entirely on CPU — slowest, minimal VRAM usage |
Outputs: upscaled_images (IMAGE)
Features:
- Automatic tiled inference via
comfy.utils.tiled_scale— no OOM on large frames - Single
movedimpre-pass avoids repeated tensor copies per frame torch.no_grad()throughout for faster inference and lower memory- Coloured progress bar with elapsed/ETA printed to the ComfyUI console
🧹 Free Video Memory
Pass-through node that explicitly flushes GPU memory between heavy pipeline stages.
| Parameter | Type | Description |
|---|---|---|
images |
IMAGE | Passed through unchanged |
aggressive_cleanup |
ENUM | Also calls torch.cuda.synchronize() and allocator cache APIs if available |
report_memory |
ENUM | Prints allocated/reserved GB before and after to the console |
Outputs: images (IMAGE) — identical to input
🔤 Text List → Batch / Text Batch → List
Bidirectional converters between ComfyUI LIST and batched STRING types, with optional delimiter joining.
⚙️ Installation
1. Clone into your ComfyUI custom nodes folder:
cd ComfyUI/custom_nodes
git clone https://github.com/FugitiveExpert01/ComfyUI-FEnodes.git
2. Restart ComfyUI.
Nodes will appear in the node menu under the FEnodes category.
No additional dependencies beyond what ComfyUI already requires (PyTorch, NumPy, Pillow).
🎬 Typical Workflows
Tiled video diffusion:
Load Video → TileSplit → [K-Sampler per tile] → TileMerge → Save Video
Video upscaling:
Load Video → Video Upscale With Model → [Free Video Memory] → Save Video
Combined:
Load Video → TileSplit → [K-Sampler per tile] → TileMerge
→ Video Upscale With Model → Free Video Memory → Save Video
🗺️ Roadmap
| Node | Description | Status |
|---|---|---|
| TileSplit | Grid tile splitting for video batches | ✅ Released |
| TileMerge | Laplacian pyramid tile reconstruction | ✅ Released |
| Video Upscale With Model | Memory-efficient per-frame upscaling | ✅ Released |
| Free Video Memory | Explicit VRAM flush between pipeline stages | ✅ Released |
| Text List → Batch | LIST to batched STRING conversion | ✅ Released |
| Text Batch → List | Batched STRING to LIST conversion | ✅ Released |
Have a node idea or a production use case that isn't covered? Open an issue.
🤝 Contributing
Pull requests are welcome. If you're adding a node, please:
- Keep video batch support (
F, H, W, C) as the primary tensor convention - Add a docstring describing what the node does and its input/output types
- Test with both single images and multi-frame video batches
- Set
CATEGORY = "FEnodes"so nodes appear grouped in the menu
📄 License
Apache License 2.0 — free to use in personal and commercial VFX pipelines.