# ComfyUI-SA-Nodes-QQ **Version 1.3.8** - 2026-04-04 - **Video Prep AB**: mid_frame_RGB changed to color and use of RGB value

⚠️ Warning: This repository was renamed in version 1.3.4. Older workflow `.json` files may still reference the previous `wanwrapper_qq` node pack, which could cause unintended installations instead of pulling this updated package.

Solution: Delete the old `wanwrapper_qq` folder from your custom_nodes directory and install the new `ComfyUI-SA-Nodes-QQ` package. Note: The recent addition of `Power Load Video` overwrites ComfyUI's default drag-and-drop behavior to enable the node creation on drag and drop. Otherwise, it should act normally with other drag-and-drop functionality. If you notice anything not working, please create an issue.

Drag and drop override to create Load video node is optional now. False by default

**Version 1.3.4** - 2026-04-01 - **ComfyUI-SA-Nodes-QQ**: Name change of node/repo to ComfyUI-SA-Nodes-QQ as Comfyui had warning print on import with kj wanwraper node pack. (Originally this was an addition node pack to build on top but now there's more nodes so it doesn't make sence as well)

![pvl](https://github.com/user-attachments/assets/e5959278-d700-401e-86e8-da5909fead56)

- **Power Load Video Node**: New node for loading and previewing videos directly in the canvas with integrated timeline playback. - **Hotkeys**: `Shift+Spacebar` to play/pause, `Shift+Drag` on `[`/`]` markers to adjust loop boundaries **Version 1.3.0** - 2026-03-16 - **Scaling Animation Update**: Enhanced box path scaling with independent h_scale (horizontal) and v_scale (vertical) controls, plus improved rotation handling that maintains visual direction during flips. **Version 1.2.8** - 2026-03-14 - Added SAM (Segment Anything Model) preparation support. Based on [MatAnyone2](https://github.com/pq-yang/MatAnyone2) [Workflow Example](examples/SAM_prep_wf.json)

![img](git_assets/img/sam.png)

https://github.com/user-attachments/assets/17440bfb-37ea-4052-9d1b-0d62d8e8b96c

**Version 1.2.1** - 2025-12-11 - Some changes to spline editor + added wan move example. Look like mix of ati and control 2.1 (update kj [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper)) [Example](examples//rabbit_control_wan_move/wan21_wan_move.json)

https://github.com/user-attachments/assets/119824e7-e309-4f00-9cc8-1404f10d91c9

https://github.com/user-attachments/assets/881f887d-d930-43d2-b197-ac44668a36e5

**Version 1.2.0** - 2025-12-01 - Added ability to connect video to bg_image and animate over it [Example](examples/ttm_ref_vid/ttm_ref_vid_example.json) - [timeline hotkey/usage](git_assets/docs/power_spline_editor.md) https://github.com/user-attachments/assets/dffe5da8-7626-45cf-ad22-777b5a14879f **Version 1.1.0** - 2025-11-26 - Added `Prepare Refs` (needs cv2) to cut out parts for `Power Spline Editor` [Example](examples/ttm_ref_cutout/ttm_ref_cutout_example.json) - `Power Spline Editor` can trigger ref preparation with Canvas button in backend without starting comfyui run

https://github.com/user-attachments/assets/7c3d1c3a-4777-432b-997e-eafa678f7e64

https://github.com/user-attachments/assets/72f46a94-a3f3-4f66-9962-c87c5c02c987

**Version 1.0.92** - 2025-11-25 - Optional color match added to `Create Image List` batch node (if no ref_image connected it's not used) - Bugfixes , Canvas refresh now can read ImageResizeKJv2 resize and match its parameters to be able to draw on same aspect ratio/crop etc. https://github.com/user-attachments/assets/e1b70e5a-71af-4d01-9034-f80c9172811f **Version 1.0.9** - 2025-11-24 - **Proper keyframes control in editor + TTM sample**: Cleaned up keyframe layer , added ref pull and animation to later be used with ttm [WF Example](examples/ttm/ttm_example_wf.json) - From https://github.com/time-to-move/TTM , ty https://github.com/kijai/ComfyUI-WanVideoWrapper for wrapper implementation. - [Power Spline Editor - keyframe hotkey usage](git_assets/docs/power_spline_editor.md) https://github.com/user-attachments/assets/41974029-4935-4952-92b6-d2d69a8873d9 **Version 1.0.8** - 2025-11-17 - **Refactor + better ATI driver control**: Slight messy refactor + proper ati point output with pause [Example](examples/ATI_buildings_wobble/ATI_buildings_wobble.json) https://github.com/user-attachments/assets/c945432c-fe58-4783-a1c3-565e61868245 **Version 1.0.6** - 2025-11-13 - **New Wan22FMLF**: A better way to add midframe control from https://github.com/wallen0322/ComfyUI-Wan22FMLF + amplitude test in one node - **Lora high.low filter**: Clicking the circle icon cycles through: ◎ → ◐ → ◑ → ◎ + right icon right click options (low/high only / both) - **Mp4 muulti cache**: save/load steps between samplers **Version 1.0.3** - 2025-11-07 - **New Hand-Drawn Spline Layer**: Added freeform path drawing with full animation controls. Special thanks to [filliptm](https://github.com/filliptm/ComfyUI_Fill-Nodes) for the implementation example - **Enhanced Spline Editor Fixes**: Improved coordinate parsing, multi-spline handling, and UI fixes across canvas, context menus, and layer rendering - **Independent Layer Point Control**: Layer points now drive independently for better animation path control --- A collection of custom nodes for ComfyUI providing video manipulation, LoRA loading, path animation, and visual drawing tools for WanVideo workflows.

https://github.com/user-attachments/assets/16f0da4c-38f1-4789-97dc-c331ad8a6d92

https://github.com/user-attachments/assets/5689c666-924a-4362-9788-d8f662c0d925

### Power Spline Editor

![img](git_assets/img/spline.png)

[Power Spline Editor - usage](git_assets/docs/power_spline_editor.md) - **Multi-layer canvas**: Draw unlimited paths with per-layer controls - **Interpolation modes**: `linear`, `cardinal` (Catmull-Rom), `basis` (B-spline), `points` (static) - **Repeat system**: Loop animations with automatic closed-path handling - **Driver/driven**: One path controls another with rotation, scaling, and smoothing - **Easing functions**: Smooth transitions with various easing curves for path animation - **Right-click controls**: Context menu for driver path with additional options - **Offset timing**: Positive/negative timing shifts with pause frame control - **Pause frames**: Start/end pause per layer for precise animation timing

### Power Lora Loader

![img](git_assets/img/lora.png)

- **LoRAs**: Dynamic UI with rgthree-style interface - **Auto-detection**: Finds Low variant pairs from High LoRAs automatically - **Smart patterns**: `-`, `_`, ` ` separators at prefix/suffix/infix positions - **Case variations**: `High`/`high`/`HIGH`/`H`/`h` → `Low`/`low`/`LOW`/`L`/`l` - **Dual strength**: Separate High (H) and Low (L) sliders with copy buttons - **Block selection**: Per-LoRA layer control with merge and memory options

### Wan Video VACE Frame Replace

![img](git_assets/img/vace.png)

[Wan Video VACE Frame Replace - example](git_assets/examples/frame_replace.json) - **Single frames**: `5` (gray), `5+` (replacement image) - **Ranges**: `22-26` (gray), `*1-5` (keep start), `1-5*` (keep end) - **a-b-c patterns**: Middle frame `b` with `a` left/`c` right context - **Multiple +**: `++a-b-c++` creates gradual duplicates on both sides - **Custom opacity**: `++a-b-c++ 0.8` for blend strength control - **Custom masks**: Auto-scaled replacement masks with opacity blending

### Wan Video Cache Samples

![img](git_assets/img/cache.png)

- **Cache/Load workflow**: Save latent samples to disk for faster workflow iteration - **Dual mode**: With input → cache and passthrough; Without input → load from cache - **Auto-path handling**: Saves to node directory with customizable cache names - **Workflow speedup**: Skip HIGH sampling steps during prompt/parameter testing to resample LOW

### Wan Video Image To Video Encode_v2

![img](git_assets/img/encode.png)

- (EXPERIMENTAL) Maybe there's native ways to do this , but it seems to work - **Multi-frame I2V**: Start/mid/end image conditioning with position control - **Temporal feathering**: Cosine interpolation for smooth frame transitions - **Latent strength**: Per-frame multipliers (start/mid/end) for motion control (default setting work but can be adjusted)

### Control only vs Control + I2V

https://github.com/user-attachments/assets/d65bb9d7-962a-4e1f-bc5d-be90adc9e311

https://github.com/user-attachments/assets/c4b70388-54a8-4b08-8f37-6028e1f2f471

### Other Experimental Node List ![Node List](git_assets/img/list.png) ## Features - **Power LoRA Loader**: Advanced LoRA loading with High/Low pair detection, block selection, and custom UI - **Power Spline Editors**: Multi-layer canvas-based path drawing with interpolation modes (linear, cardinal, basis, points) - **Video Manipulation**: Speed adjustment, frame extraction, video merging, and frame replacement - **Path Animation**: Draw shapes (circle/square/triangle) along animated paths with easing functions - **Image-to-Video Encoding**: TAEHV (Tiny AutoEncoder for Hunyuan Video) with temporal feathering and memory-efficient modes - **Cache Management**: Sample caching for faster workflow iteration - **VAE Frame Replacement**: Replace specific frames in VAE-encoded video latents ## Installation 1. Navigate to your ComfyUI custom nodes directory: ```bash cd ComfyUI/custom_nodes/ ``` 2. Clone this repository: ```bash git clone https://github.com/siraxe/ComfyUI-WanVideoWrapper_QQ.git ``` 3. Restart ComfyUI ## Requirements - ComfyUI (latest version recommended) - PyTorch with CUDA support (for GPU acceleration) - Standard ComfyUI dependencies ## Credits Experimental node package for WanVideo workflows in ComfyUI. **Special Thanks** - **KJ** - [ComfyUI-WanVideoWrapper](https://github.com/kijai/ComfyUI-WanVideoWrapper) - **rgthree** - [rgthree-comfy](https://github.com/rgthree/rgthree-comfy)

Name	Type	Description
MODEL	MODEL	—
CLIP	CLIP	—
VAE	VAE	—

Checkpoint Loader_v2 (QQ)

Inputs (0)

Outputs (3)