Comfyui-calbenodes

Comfyui-calbenodes
★ 1

角色管理LoRA应用人脸素材图像后处理
Comfyui-calbenodes 插件，包含 CharacterManagerNode、FilmGrain、FlipFlopperSameArch 节点，便于批量管理角色、应用角色专属LoRA与人脸素材、生成人脸网格并添加胶片颗粒与架构内翻转控制。
💡 在ComfyUI中管理多角色并应用专属LoRA与人脸素材。
🍴 1 Forks💻 Python🔄 2024-09-16
📦
网盘下载
复制链接后前往夸克网盘下载
https://pan.quark.cn/s/a1f1f564f19c
📄 README

CalbeNodes

A collection of custom nodes created for personal use and convenience.

Table of Contents

  • Installation
  • Usage
  • Nodes
  • Character Manager
  • Film Grain
  • Flip Flopper
  • Contributing
  • License

    • Installation

    git clone this repository into your custom nodes folder

    I tried to make it so all requirements come with comfy, so hopefully no installs needed.

    Usage

    The nodes will appear under a calbenodes heading and can be searched

    Nodes

    Character Manager

    The Character Manager node is a versatile tool for managing and applying character-specific attributes in your image generation pipeline. It allows you to create, select, and apply character settings, including LoRA models, face images, and textual descriptions.

    Features:

  • Create and manage multiple characters
  • Apply character-specific LoRA models
  • Select preferred face images for characters
  • Generate random face selections
  • Create face image grids
  • Apply character-specific activation text and descriptions

    • Inputs:

  • model: The base model to apply character settings to
  • clip: The CLIP model for text processing
  • character: Select from existing characters, create a new one, or choose randomly
  • lora_strength: Strength of the LoRA application (-10.0 to 10.0)
  • seed: Random seed for consistent results
  • new_name: Name for creating a new character
  • lora_path: Path to the character’s LoRA file
  • face_images_dir: Directory containing character face images
  • preferred_face_image: Path to the preferred face image
  • activation_text: Text to activate the character in prompts
  • description: Character description
  • negative_prompt: Negative prompt for the character

    • Outputs:

  • model: Updated model with applied LoRA
  • clip: Updated CLIP model
  • lora_activation: Character activation text
  • description: Character description
  • negative_prompt: Character-specific negative prompt
  • preferred_face: Preferred face image (as tensor)
  • random_face: Randomly selected face image (as tensor)
  • face_grid: Grid of all character face images (as tensor)
  • character_name: Name of the selected or created character
  • seed: The seed used for this execution

    • Usage:

  • Select an existing character or choose “New Character” to create one.
  • If creating a new character, provide necessary information like name, LoRA path, and face images directory.
  • Adjust the LoRA strength as needed.
  • The node will apply the character settings and return the updated model along with character-specific information and images.

    • Film Grain

    The Film Grain node adds a realistic film grain effect to images, simulating the appearance of traditional photographic film.

    Features:

  • Adds customizable film grain to images
  • Supports batch processing of multiple images
  • Adjustable grain intensity

    • Inputs:

  • image: The input image or batch of images (IMAGE type)
  • intensity: The strength of the film grain effect (FLOAT, range 0.01 to 1.0, default 0.07)

    • Outputs:

  • IMAGE: The processed image(s) with added film grain

    • Usage:

  • Connect an image or batch of images to the “image” input.
  • Adjust the “intensity” parameter to control the strength of the film grain effect.
  • The node will output the processed image(s) with the film grain applied.

    • Flip Flopper

    The Flip Flopper node (Same Architecture) is an advanced sampling node that alternates between two models during the sampling process, allowing for unique and creative image generation.

    Features:

  • Alternates between two models during sampling
  • Supports different VAEs for each model
  • Customizable sampling parameters for each model
  • Option to invert the order of model application

    • Inputs:

  • model1 and model2: The two models to alternate between
  • vae1 and vae2: VAEs corresponding to each model
  • add_noise: Enable or disable noise addition
  • noise_seed: Seed for noise generation
  • steps: Total number of sampling steps
  • cfg1 and cfg2: CFG scales for each model
  • sampler_name1 and sampler_name2: Sampler types for each model
  • scheduler1 and scheduler2: Scheduler types for each model
  • positive1, negative1, positive2, negative2: Conditioning for each model
  • latent_image: Input latent image
  • denoise: Denoising strength
  • chunks: Number of steps per chunk
  • invert: Option to invert the order of model application

    • Outputs:

  • LATENT: The resulting latent image after sampling
  • FINAL_VAE: The VAE used in the final iteration

    • Usage:

  • Connect two models, their corresponding VAEs, and other required inputs.
  • Set the sampling parameters for each model (CFG, sampler, scheduler, etc.).
  • Adjust the number of steps and chunks as needed.
  • The node will alternate between the two models during sampling, producing a unique result.

    • Contributing

    This project is primarily for personal use, but if you have any suggestions or improvements, feel free to open an issue or submit a pull request.

    License

    MIT