WanVideo Wakawave - Advanced LoRA & Prompt Tools

🌊 WanVideo Wakawave - Advanced LoRA & Prompt Tools

Advanced LoRA management and prompt building tools for WanVideo in ComfyUI.

Two powerful nodes:

• 🌊 Wakawave LoRA Loader - Dynamic LoRA management with unlimited add/remove
• 🌊 Wakawave Prompt Builder - Advanced prompt creation with segment support

---

✨ Features

Wakawave LoRA Loader

• ✅ Unlimited LoRAs - Add/remove as many LoRAs as you need
• ✅ Save/Load Presets - Save your favorite LoRA combinations
• ✅ Drag-to-Reorder - Reorder LoRAs easily
• ✅ Per-LoRA Strength - Individual strength control for each LoRA
• ✅ Enable/Disable Toggle - Test different combinations quickly
• ✅ Chainable - Connect multiple loaders together
• ✅ File Size Display - Shows individual LoRA file sizes next to each name and total combined size in header
  • Displays in node interface as a separate column
  • Shows total in header (e.g., "8 LoRAs • 6.9GB")
  • Displays in console output during execution
  • Toggle on/off with "Show File Sizes" setting
• ✅ Smart Directory Scanning - Detects all LoRA paths from extra_model_paths.yaml
• ✅ Server-Side Caching - Caches all file sizes on startup for instant display
• ✅ Compatible with WanVideo - Outputs WANVIDLORA format

Wakawave Prompt Builder

• ✅ Dynamic Prompt Management - Add/remove prompt lines on the fly
• ✅ Weight Control - Adjust emphasis for each prompt (0.5-2.0)
• ✅ Save/Load Presets - Save and recall prompt collections
• ✅ Segment Mode - Different prompts for different video segments
• ✅ Multiple Separators - Comma, newline, space, pipe, double slash, or none
• ✅ Weighted Output - Automatic formatting like (prompt:1.2)
• ✅ Chainable - Combine multiple prompt builders

---

📦 Installation

Method 1: ComfyUI Manager

1. Open ComfyUI Manager
2. Search for "WanVideo Wakawave"
3. Click Install
4. Restart ComfyUI

Method 2: Manual Installation (Recommended until comfyui manager integration)

cd ComfyUI/custom_nodes
git clone https://github.com/Wakapedia/ComfyUI-WanVideoWakawave.git
# Restart ComfyUI

Method 3: Download ZIP

1. Download the latest release
2. Extract to ComfyUI/custom_nodes/ComfyUI-WanVideoWakawave
3. Restart ComfyUI

---

🎯 Usage

Wakawave LoRA Loader

Basic Usage:

1. Add the node: Right-click → WanVideo → Loaders → Wakawave LoRA Loader
2. Click + Add LoRA to add LoRAs
3. Select LoRAs from the dropdown
4. Adjust strength sliders (0-2.0)
5. Enable/disable with checkboxes
6. Connect output to WanVideo Set LoRAs node

Save/Load Presets:

• Click 💾 Save to save current LoRA combination
• Click 📂 Load to load a saved preset
• Click 🗑 Delete to remove a preset

Chaining Multiple Loaders:

[LoRA Loader 1] → prev_lora → [LoRA Loader 2] → WanVideo Set LoRAs

File Size Display:

• Shows individual LoRA file sizes in a dedicated column (e.g., "1.3GB", "292.6MB")
• Column width dynamically adjusts based on text length
• Header displays total count and combined size (e.g., "8 LoRAs • 6.9GB")
• Toggle visibility with "Show File Sizes" in node settings
• Sizes are cached on server startup for instant display (no latency)

Example Console Output:

🌊 WanVideo Wakawave LoRA Loader
===============================================
  ✅ 1. WanAnimate_relight_lora_fp16.safetensors @ 1.00 (1.3GB)
  ✅ 2. lightx2v_I2V_14B_480p_cfg_step_distill_rank256_bf16 @ 1.00 (2.7GB)
  ✅ 3. boreal-v2.safetensors @ 0.70 (164.0MB)
===============================================
✅ Total enabled: 3 LoRAs | Combined size: 4.16GB

Wakawave Prompt Builder

Outputs:

• positive - Positive prompt text (STRING)
• negative - Negative prompt text (STRING)

⚠️ Important: The Prompt Builder outputs text (STRING), not embeddings. You must use WanVideo TextEncode Cached to convert text to embeddings before connecting to the sampler.

Basic Usage:

1. Add the node: Right-click → WanVideo → Prompts → Wakawave Prompt Builder
2. Type positive prompts in the top text box (one per line)
3. Type negative prompts in the bottom text box (optional)
4. Optional: Add weights like prompt text, weight: 1.2
5. Click + Add Line for new prompt lines
6. Connect positive → WanVideo TextEncode Cached → sampler
7. Connect negative → WanVideo TextEncode Cached → sampler

Prompt Builder Node Interface:

Visual Preset Browser:

The visual preset browser features:

• Search & Filter - Find presets instantly
• Live Preview - See prompts before loading
• Load Modes - Load both, positive only, negative only, or append
• Import/Export - Share presets as JSON files
• Rename, Duplicate, Delete - Full preset management
• Usage Tracking - See which presets you use most

Prompt Format:

nebula in deep space, weight: 1.2
vibrant purple and blue colors, weight: 1.0
stars twinkling, weight: 0.8

Or simple format (defaults to weight 1.0):

nebula in deep space
vibrant purple and blue colors
stars twinkling

Segment Mode (for 500+ frame videos):

1. Enable segment_mode
2. Format prompts with segment numbers:

0: planet Earth rotating slowly, weight: 1.0
1: camera zooming toward the moon, weight: 1.0
2: passing by asteroid belt, weight: 1.1
3: approaching distant galaxy, weight: 1.2

3. Connect a segment counter to segment_number input
4. Each segment uses its corresponding prompts

Separator Options:

• comma → prompt1, prompt2, prompt3 (default)
• newline → Prompts on separate lines
• space → prompt1 prompt2 prompt3
• pipe → prompt1 | prompt2 | prompt3
• double_slash → prompt1 // prompt2 // prompt3
• none → prompt1prompt2prompt3

---

🎬 Example Workflows

Basic LoRA + Prompt Setup

[Wakawave LoRA Loader] → lora → [WanVideo Set LoRAs] → [WanVideo Model]

[Wakawave Prompt Builder]
   ├─ positive → [WanVideo TextEncode Cached] → positive_embeds ──┐
   └─ negative → [WanVideo TextEncode Cached] → negative_embeds ──┤
                                                                    ├─→ [WanVideo Sampler]
                                                                    │
                                                 [WanVideo Model] ──┘

Multi-Segment Long Video (500+ frames)

[Segment Counter] → segment_number
                 ↓
[Wakawave Prompt Builder] (segment_mode: ON)
   ├─ positive → [WanVideo TextEncode Cached] → positive_embeds → [WanVideo Sampler]
   └─ negative → [WanVideo TextEncode Cached] → negative_embeds → [WanVideo Sampler]

Example segment prompts:
  Segment 0 (frames 0-76): "0: forest at dawn, mist rising"
  Segment 1 (frames 77-153): "1: sunlight breaking through trees"
  Segment 2 (frames 154-230): "2: birds flying through the canopy"

Multiple LoRA Combinations

[Wakawave LoRA Loader 1] (style LoRAs)
         ↓ prev_lora
[Wakawave LoRA Loader 2] (character LoRAs) → [WanVideo Set LoRAs]

---

⚙️ Settings

File Size Display & Caching

Node Setting:

• "Show File Sizes" - Toggle to show/hide file sizes in the node interface and header

Server-Side Caching:

• All LoRA file sizes are automatically scanned and cached when ComfyUI starts
• Scans all directories configured in extra_model_paths.yaml (not just the default models/loras)
• Provides instant file size display with zero latency
• Cache is stored in memory throughout the server session

Verbose Cache Output (Optional):

By default, the startup cache only shows a summary like:

[Wakawave] Scanning all configured LoRAs directories...
[Wakawave] Found LoRAs directory (from folder_names_and_paths): <your-comfyui-path>\models\loras
[Wakawave] Found LoRAs directory (from folder_names_and_paths): <extra-model-path>\loras
[Wakawave] Will scan 2 LoRAs director(ies)
[Wakawave] ✅ Cached 185 LoRA file sizes from 2 director(ies)

To see every cached file on startup, edit __init__.py:

# Set to True to show individual file caching during startup
WAKAWAVE_CACHE_VERBOSE = False  # Change to True for verbose output

When enabled, you'll see:

[Wakawave] Scanning: <your-comfyui-path>\models\loras
[Wakawave]   Cached: 'model1.safetensors' = 1326751208 bytes (1.24GB)
[Wakawave]   Cached: 'model2.safetensors' = 2929000000 bytes (2.73GB)
[Wakawave] Scanning: <extra-model-path>\loras
[Wakawave]   Cached: 'model3.safetensors' = 322519480 bytes (307.58MB)
...
[Wakawave] ✅ Cached 185 LoRA file sizes from 2 director(ies)

LoRA Loader Settings

• prev_lora (optional): Connect another LoRA loader to chain
• Hidden parameter: lora_bundle (JSON, managed by UI)

Prompt Builder Settings

• prev_prompt (optional): Previous prompt to prepend
• separator: How to join prompts (comma/newline/space/pipe/double_slash/none)
• use_weights: Enable/disable weight syntax (prompt:1.2)
• segment_mode: Enable segment-based prompting
• segment_number: Current segment number (0-100)

---

📝 Tips & Best Practices

LoRA Tips:

• Start with lower strengths (0.6-0.8) and increase gradually
• Order matters - earlier LoRAs have more influence
• Use presets to save working combinations
• Disable LoRAs to A/B test their impact
• File Sizes: Check the file size display to monitor VRAM usage
  • Larger LoRAs (500MB+) may require more VRAM
  • Use the combined total to plan your system resource allocation
  • File sizes are cached on startup for quick reference

Prompt Tips:

• Main subject: Weight 1.1-1.3
• Important details: Weight 1.0-1.2
• Background/ambient: Weight 0.7-0.9
• Stay between 0.5-1.5 for best results
• Save successful prompts as presets

Segment Mode Tips:

• Match segment frames to your context window (e.g., 77 frames)
• Plan your video story with segment markers
• Use consistent weighting within segments
• Test individual segments before full render

---

🙏 Credits

Inspired by:

• ND Super Nodes - UI framework and preset system
• WanVideo Wrapper team - Base LoRA implementation

Created by: Wakapedia

---

📄 License

MIT License - See LICENSE file for details

---

🐛 Issues & Support

Found a bug or have a feature request?

• Open an issue on GitHub
• Include your ComfyUI version and error logs

---

🔄 Changelog

v1.1.0 (Current)

• 🎉 File Size Display - Shows individual LoRA sizes in node interface as separate column
• 🎉 Smart Directory Scanning - Auto-detects all LoRA paths from extra_model_paths.yaml
• 🎉 Server-Side Caching - All file sizes cached on startup for instant display (zero latency)
• 🎉 Dynamic Column Width - File size column auto-sizes based on text length
• 🎉 Header Summary - Prominent display of total LoRAs and combined size (e.g., "8 LoRAs • 6.9GB")
• 🎉 Verbose Toggle - Optional WAKAWAVE_CACHE_VERBOSE setting to show per-file caching
• 🎉 Console Output - File sizes now display in console during execution
• ✅ Toggle Setting - "Show File Sizes" option to hide/show in node interface
• ✅ UI Improvements - Reduced widget height and optimized layout
• ✅ API Endpoint - GET /wanvideo/lora/sizes for file size queries
• ✅ Fallback Support - Automatically calculates sizes if not in cache

v1.0.5

• Previous stable release

v1.0.0 (Initial Release)

• 🌊 Wakawave LoRA Loader with unlimited LoRAs
• 🌊 Wakawave Prompt Builder with segment support
• ✅ Save/load presets for both nodes
• ✅ Drag-to-reorder functionality
• ✅ Weight control for prompts and LoRAs
• ✅ Multiple separator options
• ✅ Chainable nodes

---

Enjoy creating amazing videos with Wakawave! 🌊