### [Detailed manual for included nodes](Workflow/Manual/nodes/minimal_workflow.md)
+Core generation pipeline. Single prompt input with full model concept support.
+
+**Supported model concepts:** SD1, SD2, SDXL, Illustrious, SD3, StableCascade, Chroma, Z-Image, Turbo, Flux, Nunchaku, QwenGen, QwenEdit, WanImg, KwaiKolors, Hunyuan, Playground, Pony, LCM, Lightning, Hyper, PixartSigma, SANA1024, SANA512, AuraFlow. Future support: HiDream, Mochi, WanT2V, WanI2V, Cosmos, Flux2, SSD, SegmindVega, KOALA, StableZero, SV3D, SD09, StableAudio, LTXV.
+
### Minimal workflow features:
-
+
+- Central model concept selector node controls sampler, VAE, CLIP settings per model type
+- Automatic model keyword insertion to prompt
- Prompt selector to any prompt sources
-- Prompt can be saved to `CSV` file directly from the prompt input nodes
+- Prompt can be saved to `CSV` file directly from the prompt input nodes
- `CSV` and `TOML` file readers for saved prompts, automatically organized, saved prompt selection by preview image (if preview created)
- Randomized latent noise for variations
- Prompt encoder with selectable custom clip model, long-clip mode with custom models, advanced encoding, injectable internal styles, last-layer options
-- Sampler with `variation extender` and `Align Your Step` features
-- A1111 style network injection supported by text prompt (Lora, Lycorys, Hypernetwork, Embedding)
-- Automatized and manual image saver. Manual image saver with optional **preview saver** for checkpoint (Lora, Lycoris, Embedding) selectors and saved .csv prompts
+- Sampler with `variation extender` and `Align Your Steps` features
+- A1111 style network injection supported by text prompt (LoRA, LyCoris, Hypernetwork, Embedding)
+- Automatized and manual image saver with optional **preview saver** for checkpoint selectors and saved .csv prompts
+- Aesthetic scorer for final image quality assessment
- Upscaler (selectable Ultimate SD and hiresFix)
- Dynamic prompt support
-- Auto clean incompatible network tags from prompt by model arhitechture
+- Auto clean incompatible network tags from prompt by model architecture
### [Detailed manual for included nodes](Workflow/Manual/nodes/basic_workflow.md)
+Professional prompt development workflow. Extended prompt management for testing and iteration.
+
+**Same model support as Minimal workflow.**
+
### Basic workflow features:
#### Same as Minimal workflow plus:
-- **Half-automatic model concept selector:**
- - **Supported concepts:** SD1, SD2, SDXL, SD3, StableCascade, Turbo, Flux, KwaiKolors, Hunyuan DiT (image only), Playground, Pony, LCM, Lightning, Hyper, PixartSigma, Sana (both 1024 and 512)
- - Custom (and different) sampler settings for all concepts. The main idea is set sampler nodes only one time (`sampler`, `scheduler`, `step`, `cfg`) then just select model only what will use right sampler, vae, clip settings by `Model concept selector`.
- - Auto detection of selected model type (if data already stored on external .json file, see longer [manual](Workflow/Manual/nodes/basic_workflow.md))
- - Auto **download** and apply Hyper, Lightning, and Turbo speed loras at first usage from here: https://huggingface.co/ByteDance/Hyper-SD/tree/main **check your SSD space before!**
-
-**On the `Concept selector` node you will see `None` on all required fields, for example on Cascade and Flux files. You must install all required additional files manually to right path, then select correct model/clip/vae on lists.**
+- 12 prompt inputs with 1-click selector for prompt switching
+- Efficient workflow for prompt developers testing multiple variations
### [Detailed manual for included nodes](Workflow/Manual/nodes/basic_production_workflow.md)
-### Basic production workflow features:
+Full production pipeline with styling, refinement, and selective output.
+
+**Same model support as Minimal workflow.**
+
+### Basic Production workflow features:
#### Same as Basic workflow plus:
-- Added 4 test and 4 development prompt inputs, easy to switch
-- Local LLM models can help refine/repair prompts. Refined prompts can be added to original, replace original, or keep original as L prompt but send refined to T5-XXL clip is avalable in clip encoder node
-- Customizable refiner blocks for face, eye, mouth, and hand refining. Auto segmentation model downloads deleted, manual model download required
-- Refiner blocks using [DeepFace analyzer](https://github.com/serengil/deepface) if needed, detect age, race, gender and mood
+- 19 prompt sources for special cases (Daily Challenges, Articles, etc.)
+- Style block: Style Pile, Midjourney, Emotions, Camera Lens (prompt injection styling)
+- 4 separated refiner detailer blocks: Face, Eye, Mouth, Hands
+- Selective image saver: saves only images exceeding user-defined aesthetic score threshold (portfolio filtering)
-
-### [Detailed manual for included nodes](Workflow/Manual/nodes/basic_production_plus_workflow.md)
-
-### Basic production plus workflow features:
-
-#### Same as Basic production workflow plus:
-
-- Post-processing nodes, have to install [Image magick](https://imagemagick.org/index.php) and [MagickWand nodepack](https://github.com/Fannovel16/ComfyUI-MagickWand)
-- Several style nodes including `Midjourney style`, `Camera lenses`, `Style pile`, and `Emotions` at `Style block` group
-- 2 prompt inputs for DailyChallenges of AI image communities
-- Several refinements on nodes
\ No newline at end of file
+
-
-The Visual Checkpoint Selector helps you choose and manage AI models (checkpoints) in your workflow with a visual (or legacy list) interface.
-
-### Basic Usage:
-Select your AI model from the `base_model` dropdown list, which shows all installed models on your system.
-
-### Visual Selection Mode:
-Toggle `show_modal` to ON to open a visual gallery of preview images for each model. This makes it easier to choose the right model by seeing example outputs.
-
-### Preview Settings:
-- Select `preview_path`:
- - `Primere legacy` - uses custom path for preview images
- - `Model path` - uses preview images from your original model folder
-
-- `show_hidden` controls visibility of hidden files/folders (those starting with a dot)
-
-Visual `checkpoint` selection, automatized filtering by subdirectories (first row of buttons) and versions (second row of buttons):
-
-
-
-### Random Model Feature:
-Turn on `random_model` to automatically select random models from the current folder. For example, if you've selected checkpoint from "Photo" folder, it will randomly pick from any model in that folder. This is useful for batch processing too.
-
-### Aesthetic Score percent display:
-- `aescore_percent_min`: Because the preview images show aesthetic scores if saved and measured, this value or less mean 0%
-- `aescore_percent_max`: Because the preview images show aesthetic scores if saved and measured, this value or more mean 100%
-
-These scores help sorting models based on their quality ratings.
-
-**When all data available, thse badges will visible in the preview:**
-
-
-
-- **Top left:** model concept (Flux, SD1, SD2, SDXL, etc...)
-- **Top right:** if symlinked, what type of diffuser linked
-- **Botom:** the average aesthetic score. Have to use aesthetic scorer node before store this data for checkpoints and saved prompts. The number is the average, but the percent depending on the checkpoint selector settings, where the `aescore_percent_min` and lower value mean 0%, `aescore_percent_max` and higher mean 100%.
+# Minimal Workflow - Node Groups Manual
+The Dashboard consolidates model loading, seed control, and resolution settings in a unified interface. This group handles all baseline generation parameters before prompt encoding.
-This node helps you set the perfect image dimensions for your generations with preset ratios or custom settings.
-
-### Basic Resolution Selection:
-- Choose from predefined aspect ratios in the `ratio` dropdown (Photo, Portrait, Old TV, HD, HD+, Square, etc.)
-- Use `resolution` mode:
- * "Auto" - automatically sets base resolution based on your selected model. If the model version available this settings useful.
- * "Manual" - lets you input custom dimensions bases using:
- - `sd1_res`: Base resolution for SD1 models (768 default)
- - `sdxl_res`: Base resolution for SDXL models (1024 default)
- - `turbo_res`: Base resolution for Turbo models (512 default)
-
-### Image Orientation:
-- Set `orientation` to Horizontal or Vertical
-- Enable `rnd_orientation` to randomly switch between orientations. This function useful for batch generation.
-- `round_to_standard` re-count dimensions to "standard" of selected AI models
-
-### Custom Ratio Settings:
-Enable `calculate_by_custom` to use your own aspect ratios (example: 1.6:2.8):
-- `custom_side_a`: First side ratio (e.g., 1.60)
-- `custom_side_b`: Second side ratio (e.g., 2.80)
-
-Note: Aspect ratios can be customized by editing the external .toml configuration file from path: `Toml/resolution_ratios.toml`
+
+**Purpose:** Select and load AI model checkpoints with visual preview interface, automatic directory filtering, and aesthetic quality scoring.
-The Prompt node provides advanced prompt control with organization features and special settings.
+#### Primary Settings:
-### Prompt Inputs:
-- `positive_prompt`: Enter your main prompt describing what you want to create
-- `negative_prompt`: Enter elements you want to avoid in the generation
+| Setting | Purpose |
+|---------|---------|
+| `base_model` | Select checkpoint from dropdown or visual gallery |
+| `show_modal` | Toggle visual gallery preview mode ON/OFF |
+| `preview_path` | Choose preview image source: "Primiere legacy" or "Model path" |
+| `show_hidden` | Show/hide hidden files and folders (dot-prefixed) |
+| `random_model` | Automatically select random checkpoint from current folder |
-### Organization Features:
-- `subpath`: Save your generated images into themed folders (e.g., "CutePets", "Sci-Fi, etc...")
-- `model`: Choose a specialized model for specific types of images (e.g., interior, exterior, etc...). This is standard model list.
+#### Aesthetic Scoring Display on visual previews:
-### Orientation Control:
-Choose how to handle image orientation:
-- `None`: Use default orientation from `Resolution Selector` node
-- `Random`: Randomly switch between **horizontal** and **vertical**. This function useful for batch generation.
-- `Horizontal`: Force **horizontal** composition
-- `Vertical`: Force **vertical** composition
+| Setting | Purpose |
+|---------|---------|
+| `aescore_percent_min` | Lower bound for quality scaling (maps to 0%) |
+| `aescore_percent_max` | Upper bound for quality scaling (maps to 100%) |
-The orientation setting helps compose your image properly for your selected subject matter.
+### The visual preview modal:
-### Save prompt to external file:
-With `Save prompt to file...` button you can save current prompt to external CSV file, the `Visual Style Selector` and `Visual Prompt CSV` nodes will read by name.
-Read the manual of **Visual Prompts (style) Selector** and **Visual Prompts - auto organized** nodes later.
-- To the prompt saver dialog you can enter or choose name for prompt. If you choose existing prompt name from the `Prompt name` list, the original prompt with same name will be overwritten
-- You can select existing or create new category what will be use as folder name and category for generated image
-- You can edit both positive and nagetive prompt, but positive prompt required
-- The `Preferred model` and `Preferred orientation` inputs are read only. These data must be set on the prompt input node before open the save dialog
-- Cancel or finish prompt by save buttons
-- Make backup of your previous `styles.csv` file on the `custom_nodes\ComfyUI_Primere_Nodes\stylecsv\` folder before use prompt saver
+
-
+#### Interface Layout:
-[This video contains tutorial for prompt engineering](https://www.youtube.com/watch?v=joqvC8vb6Xo)
+**Row 1 - Directory Structure:** Filter checkpoints by folder organization (e.g., Root, Flux, SD1, SDXL, Photo, Design, Character, Style, etc.). Buttons represent your checkpoint folder hierarchy for quick categorization.
+
+**Row 2 - Model Types:** Filter by supported model concept `(SD1, SD2, SDXL, Flux, Hunyuan, LCM, Lightning, Playground, Pony, SD3, Turbo, StableCascade, KwaiKolors, PixartSigma, SANA, Illustrious, Nunchaku, OwnGen, Wan2V, Wan1Edit, AuraFlow, Chroma, Waning, HiDream, Mochi, Cosmos, Flux2, LTXV, Z-Image)`.
+
+**Row 3 - Controls:** Filter text search, clear filter, sort options `(aScore, Name, Version, Path, Date, Symlink, STime)`, and sort direction.
+
+#### Preview badges show:
+
+
+
+- **Top left:** Model concept (Flux, SD1, SDXL, etc.)
+- **Top right:** Symlink type if applicable
+- **Bottom:** Average aesthetic score as percentage (based on min/max range)
+
+**Note:** Aesthetic scores require pre-computed data from aesthetic scorer node stored with checkpoint.
+
+---
+
+### Checkpoint Loader
+
+Automatically loads selected checkpoint, VAE, and CLIP model based on Visual Checkpoint Selector choice and automatically deteted model concept. This node ccontolled by the data tupple on the external `control_data` input. Outputs:
+- `loaded_model`: Model tensor
+- `loaded_clip`: CLIP encoder
+- `loaded_vae`: VAE decoder
+- `control_data`: Model metadata and version info
+
+**No manual configuration required** — uses cached settings of saved model concept json.
+
+---
+
+### Fast Seed Control
+
+`Why fast? Because no fron-tend for seed generation, only for result display. For large queue settings much faster than anything else.`
+
+| Setting | Purpose |
+|---------|---------|
+| `seed_setup` | "Random" = new seed each run, "Custom" = use fixed value |
+| `custom_seed` | Fixed seed value when `seed_setup` = "Custom" |
+| `random_seed` | Read-only output showing last generated seed |
+
+---
+
+### Resolution Selector
+
+**Purpose:** Define output image dimensions with preset aspect ratios or custom settings.
+
+#### Basic Selection:
+
+| Setting | Options | Purpose |
+|---------|------------------------------------------------|---------|
+| `ratio` | Photo, Portrait, Square, HD, HD+, Old TV, etc. | Predefined aspect ratios |
+| `resolution` | Auto / Manual | "Auto" = model-based defaults, "Manual" = custom base resolution |
+
+#### Model-Based Auto Resolution:
+
+When `resolution` = "Auto":
+- resolution automatically set by the internal settings for actual selected model concept.
+
+#### Manual Resolution:
+
+When `resolution` = "Manual", manually specify base dimensions per model:
+- set the target resolution in megapixels on `manual_res` combo. This will overwrite the automatic value.
+
+#### Orientation Control:
+
+| Setting | Purpose |
+|---------|---------|
+| `orientation` | Horizontal or Vertical composition |
+| `rnd_orientation` | Randomly alternate between orientations (useful for batch generation) |
+| `round_to_standard` | Force dimensions to model-native standards |
+
+#### Custom Aspect Ratios:
+
+Enable `calculate_by_custom` to define custom ratios (e.g., 1.6:2.8):
+- `custom_side_a`: First ratio component (e.g., 1.60)
+- `custom_side_b`: Second ratio component (e.g., 2.80)
+
+**Note:** Preset ratios editable in `Toml/resolution_ratios.toml`
+The Prompt Input Group handles manual prompt entry, saving, and visual selection of pre-built or saved prompts. This group provides flexible workflows for both real-time prompt writing and reusable prompt library management.
-A control node that lets you quickly switch between different prompt sources in your workflow, including `Style Selector` nodes.
-
-### How It Works:
-- Connect multiple prompt sources to this node (any different sources), but the connection and unconnection queue is important. If failed, just reload the browser.
-- Use the `select` input to choose which prompt source to use (1-any)
-- The selected prompt source becomes active, while others remain inactive
-
-### Usage Example:
-If you have different prompts for:
-- Portrait shots
-- Landscape scenes
-- Character designs
-
-You can connect all of them to the Prompt Switch and easily toggle between them using the selector, without needing to reconnect nodes or modify your workflow.
+
+**Purpose:** Write positive and negative prompts directly with organization metadata, then save to external CSV file for later reuse.
-A powerful tool that lets you save and load complete prompt configurations using a visual interface or simple list.
+#### Inputs:
-### Basic Usage:
-- Choose from saved styles using the `styles` dropdown
-- Toggle `show_modal` to switch between:
- * List view: Simple dropdown of style names
- * Visual view: Preview images of each style's output
+| Input | Purpose |
+|-------|---------|
+| `PROMPT+` | Positive prompt text area (multiline) |
+| `PROMPT-` | Negative prompt text area (multiline) |
+| `SUBPATH` | Folder organization category for saved images |
+| `MODEL` | Preferred checkpoint for this prompt |
+| `ORIENTATION` | Preferred image orientation (None/Horizontal/Vertical/Random) |
-Visual `saved prompt` selection `(csv source)`, automatized filtering by categories:
+#### Outputs:
-
+| Output | Purpose |
+|--------|---------------------------------------------------------|
+| `PROMPT+` | Positive prompt text |
+| `PROMPT-` | Negative prompt text |
+| `SUBPATH` | Image save category |
+| `MODEL` | Preferred model |
+| `ORIENTATION` | Preferred orientation |
+| `PREFERRED` | Aggregation of model, subpath + orientation preferences |
-### Style Components:
-Each saved style can include:
-- Name of saved prompt
-- Positive and negative prompts
-- Specific orientation
-- Custom save folder (subpath)
-- Model preference
+#### Save to External CSV:
-### Control Options:
-- `use_subpath`: Apply the style's saved folder path
-- `use_model`: Use the style's recommended/preferred model
-- `use_orientation`: Apply the style's preferred orientation
-- `show_hidden`: Show/hide styles names or path contains word `nsfw`
-- `random_prompt`: Randomly select prompt from available styles from same subpath as selected
+Click "Save prompt to file..." button to open the prompt saver dialog. This saves your manually crafted prompt to an external CSV file for reuse via Visual Style Selector or Visual Prompt CSV nodes.
-### Quality sorting:
-- `aescore_percent_min`: Because the preview images show aesthetic scores if saved and measured, this value or less mean 0%
-- `aescore_percent_max`: Because the preview images show aesthetic scores if saved and measured, this value or more mean 100%
+
-Note: Styles are stored in an external .csv file that can be easily edited and shared. Rename `stylecsv/styles.example.csv` to `stylecsv/styles.csv` and use your own prompt collection.
+**Dialog Fields:**
+- **Prompt name:** Create new or overwrite existing saved prompt from list
+- **Prompt category (subpath):** Assign category/folder (e.g., "SeasonBackground", "Architecture", "Nature")
+- **Positive prompt:** Auto-populated from input, editable
+- **Negative prompt:** Auto-populated from input, editable
+- **Preferred Model:** Read-only, set from Primiere Prompt input node
+- **Preferred Orientation:** Read-only, set from Primiere Prompt input node
-
+---
-This node organizes your saved prompts into categories for easier management, especially useful when you have a large saved collection of prompts.
+### Visual Style Selector
-### Category Organization:
-- Prompts are automatically sorted into categories like:
- - Architecture
- - Art
- - Character
- - ...and more dependign your source .csv.
+**Purpose:** Select and load pre-saved prompts with visual gallery preview and category filtering.
-### How to Use:
-1. Select a category from node
-2. Choose a specific prompt from that category
-3. All related settings (prompt, model, path) load automatically from .csv file: Rename `stylecsv/styles.example.csv` to `stylecsv/styles.csv` and use your own prompt collection.
-
-## Control Options (same as than the Visual Prompts (style) Selector node)
-- `show_modal`: Switch between list and visual preview mode
-- `show_hidden`: Include/ignore hidden category items if name or path contains word `nsfw`
-- `use_subpath`: Use saved folder paths (usually as prompt category)
-- `use_model`: Apply recommended models
-- `use_orientation`: Use saved orientation settings
-- `random_prompt`: Pick random prompt from selected category
-
-## Benefits
-- Organized browsing instead of one long list
-- Quick access to themed prompts
-- Easy to find related styles
-- Categories are created automatically from your saved paths
-
-Note: Categories are created from the folder structure in your styles.csv file, making organization automatic and maintenance-free.
-
-
-
-This node processes dynamic text prompts with random variations and maintains seed control for consistent results. The node has two inputs: a dynamic prompt string and a seed value.
-
-### Input Fields:
-- `dyn_prompt`: Takes your prompt text containing dynamic sections
-- `seed`: Controls randomization for consistent results
-
-### Usage:
-This node helps create varied prompts while maintaining reproducibility. When you input a prompt with dynamic sections (like {red|blue|green}), the node will randomly select one option based on the seed value. Perfect for batch processing or exploring variations while keeping track of successful combinations.
-
-### Benefits:
-- Consistent randomization with seed control
-- Streamlines prompt variation workflow
-- Integrates seamlessly with other prompt processing nodes
-- Reduces manual prompt editing time
-- Perfect for batch generation with controlled variations
-
-Read manual of dynamic prompt syntax: https://github.com/adieyal/sd-dynamic-prompts/blob/main/docs/SYNTAX.md
-
-
-
-This KSampler node extends the standard sampling functionality with fine-tuned variation controls and performance options.
-
-#### Variation System:
-- `variation_extender`: Fine-tunes noise injection from 0.0 to 1.0, allowing subtle variations while maintaining the original image's core elements
-- `variation_batch_step`: Enables progressive variation in batch processing by incrementing noise injection per step (e.g., 0.1 increment over 10 steps creates a gradual transformation sequence)
-- `variation_level`: When set to "Maximize", randomizes the noise injection values for more diverse outputs
-
-#### Performance Options
-- `device`: Select processing hardware (CPU/GPU/Default)
-- `align_your_steps`: Implements NVIDIA's AlignYourSteps technology, which helps maintain temporal consistency and reduces unwanted artifacts during the sampling process. Read details from here: https://research.nvidia.com/labs/toronto-ai/AlignYourSteps/
-
-### Additional Settings:
-- `denoise`: Controls the denoising strength
-- `model_sampling`: Adjusts the model's sampling parameters (usually for SD3 models only)
-
-
-
-This sophisticated prompt encoder node offers extensive control over prompt processing with multiple CLIP and LONG-CLIP models and advanced encoding options.
-
-### CLIP Processing:
-- `clip_mode`: Toggle between standard CLIP and Long-CLIP processing
- - `clip_model/longclip_model`: Model selection based on clip_mode switch
-- `last_layer`: Fine-tune CLIP encoding by selecting specific negative layers for feature extraction
-- `negative_strength`: Adjusts the intensity of negative prompt influence (if the selected model support)
-
-### Style System:
-- `use_int_style`: Enables internal style system loaded from external .toml configurations from the path: `Toml/default_neg.toml` and `Toml/default_pos.toml`
- - `int_style_pos/int_style_neg`: Select predefined styles by name
- - `int_style_pos_strengt/int_style_neg_strength`: Control strength of applied styles
-
-### Advanced Encoding Options:
-- `adv_encode`: Enables alternative (advanced) CLIP encoding methodology
-- `token_normalization`: Controls how token weights are normalized (mean/none/length/length+mean)
-- `weight_interpretation`: Defines how prompt weights are processed (comfy++/A1111/comfy)
-
-### Enhanced Prompt System:
-- `enhanced_prompt_usage`: Controls enhanced prompt processing
- - `None`: Ignores enhanced prompt
- - `Add`: Appends to end of positive prompt
- - `Replace`: Replace positive prompt (very strong modification)
- - `T5-XXL`: Uses enhanced prompt input for T5-XXL encoding if concept support T5 clipping
-- `enhanced_prompt_strength`: Controls enhanced prompt influence if added to original positive prompt
-
-### Additional Style Controls:
-- `style_handling`: Separate or merge styles with original prompt
-#### If style and prompt merged:
-- `style_position`: Placement of additional style (Top/Bottom)
-#### If style and prompt separated:
-- `style_swap`: Style send to default clip, prompt send to T5/L encoder or Style send to T5/L, prompt send to default clip
-
-### Optional prompt handling:
-- `opt_pos_strength/opt_neg_strength`: Fine-tune optional prompt strengths
-- `l_strength`: Controls SDXL first-pass prompt intensity id L prompt connected to node
-
-
-
-This node evaluates the aesthetic quality of generated images and saving scoring data to display in visual previews. It provides numerical scoring and statistical tracking for your generations.
-
-**The required model files must be downloaded manually. Download all source files to the target path:**
-- `model 1`: **cafe_aesthetic** Source: https://huggingface.co/cafeai/cafe_aesthetic/tree/main target: {your_comfy_model_folder}\aesthetic\cafe_aesthetic\
-- `model 2`: **cafe_style** Source: https://huggingface.co/cafeai/cafe_style/tree/main target: target: {your_comfy_model_folder}\aesthetic\cafe_style\
-
-
-### Key Features:
-- `get_aesthetic_score`: Enables/disables image quality scoring
- - `add_to_checkpoint`: Save scores to checkpoint data, allowing you to sort image quality of different model checkpoints
- - `add_to_saved_prompt`: Save scores to saved prompts, helping identify (and sort) consistently high-performing prompts
-
-### Benefits:
-- Track generation quality automatically
-- Compare performance across different checkpoints and prompts
-- Identify your most successful prompts through statistical tracking
-- View average scores in visual selectors
-- Make data-driven decisions about your workflow settings
-
-The scoring system helps optimize your workflow by providing objective (or subjective?) feedback on image quality and maintaining statistics for both checkpoints and prompts.
-
-
-
-This versatile node provides comprehensive image saving functionality with **visual preview management** capabilities for your workflow.
-
-### Save Modes:
-#### Preview Save Mode: `Save as preview`
-- Saves images as visual previews for checkpoints, LoRAs, Lycoris, Hypernetworks, Embedding, and saved prompt selections
-- `Overwrite`: Replace existing or create new preview
-- `Keep`: Preserve existing, only create if missing
-- `Join horizontal`: Combine horizontally with existing preview or create new
-- `Join vertical`: Combine vertically with existing preview or create new
-- `Target selection`: Select only one target if more than one available in the process, for example Loras or Embeddings.
-
-#### The bonus hidden feature, that one click very close under the save button, the previously saved preview visible if exist.
-
-
-
-#### Local Storage Mode: `Save as any`
-- `Format Options`: PNG, JPEG, WebP
-- `Size Control`: Resize by specifying maximum dimension while preserving aspect ratio. 0 mean no resize
-- `Quality Settings`: Adjust compression for JPEG/WebP formats
-
-### Benefits:
-- Create visual reference libraries for models, additional networks and prompts
-- Flexible preview management for workflow organization
-- Custom export settings for different use cases
-- Space-efficient preview combinations
-- Maintain organized model and prompt libraries with visual references
-
-
-
-This node allows you to set a target resolution for upscaling images while preserving the original aspect ratio. Below are the available settings:
-
-- `image`: Connect the input image
-- `width & height`: Connect the original dimensions of the image
-
-### Settings:
-
-- `use_multiplier`: Enable or disable the upscaler
-- `upscale_to_mpx`: Specify the final target resolution in megapixels, maintaining the original aspect ratio
-- `triggered_prescale`: Pre-resizes the image if it’s smaller than the value in **area_trigger_mpx**. This setting helps reduce resource demand for extreme upscaling
-- `area_trigger_mpx`: The megapixel size threshold to trigger prescaling. Useful for minimizing resource consumption
-- `area_target_mpx`: Sets the target size (in megapixels) before sending the image to the upscaler. It is recommended to set this to about a quarter to half of the final target size
-- `upscale_model`: Select an upscaling model from the standard list of available options
-- `upscale_method`: Method for resizing the image if **triggered_prescale** is activated (as shown in the screenshot)
-
-### Benefits:
-
-- `Efficient Resource Management`: Reduces resource demand by prescaling smaller images before upscaling if the difference between source and target size too much
-- `Flexible Settings`: Various settings to customize the upscaling process according to your needs
-
-
-
-This node automatically saves images along with metadata, generated by the full workflow, to specified paths. Below are the settings available for configuring the node:
-
-- `images`: Connect the input image
-- `image_metadata`: Connect the metadata generated by the workflow
+
#### Settings:
-- `save_image`: Enable or disable the saving of images
-- `aesthetic_trigger`: Set a minimum required aesthetic score. If set to 0, all images will be saved. If set to a value greater than 0, only images meeting this score threshold are saved. If no score measure, all images will be saved
-- `output_path`: Specify the final image path using Python-style format. My favorite setting is: `{my_local_path}\[time(%Y-%m-Week-%W)]` for weekly subdirectories
-- `subpath`: Select a subpath based on the project type (e.g., dev, test, fashion, etc...)
-- `subpath_priority`: Choose either "Selected subpath" uses `subpath` value or set to "Preferred" uses workflow metadata settings for example from simple `Prompt` node or `Saved csv prompts` node
-- `add_modelname_to_path`: If enabled, includes the checkpoint name as a subfolder
-- `add_concept_to_path`: If enabled, includes the model concept name (e.g., SD1, SD2, SDXL, Flux) as a subfolder
-- `filename_prefix`: Set the prefix for the target filename
-- `filename_delimiter`: Specify the character used between filename parts
-- `add_date_to_filename`, `add_time_to_filename`, `add_seed_to_filename`, `add_size_to_filename`, `add_ascore_to_filename`: Boolean switches to include date, time, seed, size, or aesthetic score in the filename
-- `filename_number_padding`: Number of padding digits if duplicate filenames exist
-- `extension`: Set the image file format (e.g., jpg, png, webp, etx)
-- `png_embed_workflow`: If enabled, saves the full workflow to the PNG file for loading to ComfyUI
-- `png_embed_data`: Special data for Premiere Nodepack’s `Image Recycler` node to read back generation details
-- `image_embed_exif`: Saves generation data to JPEG EXIF for `Image Recycler` node
-- `a1111_civitai_meta`: Saves A1111-compatible metadata auto-readable by Civitai portal
-- `quality`: Set quality for jpg and webp images
-- `overwrite_mode`: If enabled, overwrites existing files
-- `save_meta_to_json`: Saves generation details to an external `.json` file
-- `save_meta_to_txt`: Saves generation details to an external `.txt` file
+| Setting | Options | Purpose |
+|---------|---------|---------|
+| `styles` | Dropdown | Choose saved prompt from list |
+| `show_modal` | ON/OFF | Toggle between dropdown list and visual gallery preview |
+| `show_hidden` | ON/OFF | Show/hide items with "nsfw" in name or path |
+| `use_subpath` | ON/OFF | Apply the saved prompt's folder category to output |
+| `use_model` | ON/OFF | Apply the saved prompt's preferred checkpoint |
+| `use_orientation` | ON/OFF | Apply the saved prompt's preferred orientation |
+| `random_prompt` | ON/OFF | Randomly select from same category as currently selected prompt |
+| `aescore_percent_min` | Integer | Lower quality threshold (maps to 0%) for preview sorting |
+| `aescore_percent_max` | Integer | Upper quality threshold (maps to 100%) for preview sorting |
-### Benefits:
+#### Modal Interface:
-- `Automated Saving`: Simplifies saving images and metadata with configurable filenames and paths
-- `Detailed Metadata`: Saves essential generation details for later review or sharing
-- `Compatibility`: Compatible with tools like `Image Recycler` and Civitai for enhanced workflow integration
-- `Customizable File Management`: Options to manage paths, subfolders, and filenames for organized storage
+**Row 1 - Category Buttons:** Filter by saved prompt categories `(Architecture, Art, Character, Design, Edit, Horror, Influencer, Nature, Photography, Sci-Fi, Vehicles, Others)`
-
+**CSV Source:** Rename `stylecsv/styles.example.csv` to `stylecsv/styles.csv` and populate with your own prompts. Format: `columns for name, positive, negative, category (subpath), preferred model, preferred orientation`.
-### Overview:
-The Network Tag Cleaner node provides intelligent management of network adapter tags in prompts, supporting both manual and automatic cleaning modes for various model architectures.
+---
-### Supported Network Types:
-- Embeddings (`embedding:name`)
-- LoRA (`- -# Network Tag Loader: - -
-
-This node manages additional networks such as Lora, Lycoris, and hypernetworks from prompt in A1111 style like: `
+
+**Requirement:** Embedding file must exist with exact name match.
+
+#### Inputs:
+
+| Input | Purpose |
+|-------|---------|
+| `positive_prompt` | Positive prompt with potential embedding keywords |
+| `negative_prompt` | Negative prompt with potential embedding keywords |
+
+#### Outputs:
+
+| Output | Purpose |
+|--------|---------|
+| `positive_prompt` | Converted positive prompt with embedding: prefixes |
+| `negative_prompt` | Converted negative prompt with embedding: prefixes |
+
+---
-# Primere Embedding Handler: +## Encoder, Sampler, Decoder Group -
+The Encoder, Sampler, Decoder group handles prompt encoding to latent space, noise generation, sampling/diffusion, latent-to-image conversion, and quality scoring. This is the core generation pipeline.
-A ComfyUI node that automatically converts Automatic1111-style textual inversion embedding syntax to ComfyUI-compatible format.
+
-### Description:
+-This node allows you to use Automatic1111-style embedding syntax in your prompts while working in ComfyUI. It automatically detects embedding keywords and converts them to the proper ComfyUI format. +### Primiere Noise Latent -### How It Works: +**Purpose:** Generate random latent noise tensor as initialization for diffusion process. -#### Input Format (Automatic1111 style): -- Simple keywords like `Autumn` or `Summer` -- No special prefix needed -- Must match exactly with embedding filename (case-sensitive) without extension +#### Inputs: -#### Output Format (ComfyUI style): -- Converted to `embedding:Autumn` or `embedding:Summer` -- Properly formatted for ComfyUI processing -- Only converts if matching embedding file exists +| Input | Purpose | +|-------|---------| +| `width` | Image width in pixels | +| `height` | Image height in pixels | +| `noise_seed` | Seed for reproducible noise generation | +| `optional_vae` | VAE for noise scaling (optional) | -### Example: +#### Settings: -Input prompt: -``` -This prompt contains Autumn. The embedding name is case sensitive: summer just a word, but Summer is embedding. -``` +| Setting | Purpose | +|---------|----------------------------------------------------| +| `rand_noise_type` | Randomize noise type per generation (ON/OFF) | +| `noise_type` | Noise distribution: white, gaussian, etc. | +| `rand_alpha_exponent` | Randomize alpha exponent per generation (ON/OFF) | +| `alpha_exponent` | Alpha blending exponent (default 1.000) | +| `alpha_exp_rand_min` | Minimum random alpha exponent (default 0.500) | +| `alpha_exp_rand_max` | Maximum random alpha exponent (default 1.500) | +| `rand_modulator` | Randomize modulation per generation (ON/OFF) | +| `modulator` | Noise modulation factor (default 1.00) | +| `modulator_rand_min` | Minimum random modulation (default 0.80) | +| `modulator_rand_max` | Maximum random modulation (default 1.40) | +| `rand_device` | Randomize device (GPU/CPU) per generation (ON/OFF) | +| `device` | Compute device: cpu, cuda | +| `expand_random_limits` | Expand limits beyond defaults (ON/OFF) | -Output prompt: -``` -This prompt contains embedding:Autumn. The embedding name is case sensitive: summer just a word, but embedding:Summer is embedding. -``` +#### Outputs: -### Features: +| Output | Purpose | +|--------|---------| +| `LATENTS` | Generated latent noise tensor | +| `PREVIEWS` | Preview images of noise pattern | +| `CONTROL_DATA` | Metadata and generation settings | -- Seamless conversion between Automatic1111 and ComfyUI embedding formats -- Case-sensitive matching to prevent unwanted conversions -- Only converts existing embeddings -- Works with both positive and negative prompts -- Real-time conversion during workflow execution +--- -### Benefits: +### Primiere Prompt Encoder -- Use Automatic1111 prompts directly in ComfyUI without manual conversion -- Maintain compatibility with existing prompt collections -- Reduce errors from incorrect embedding syntax -- Save time on prompt formatting +**Purpose:** Encode positive and negative prompts into CLIP embeddings for guidance during sampling. -
\ No newline at end of file +#### Inputs: + +| Input | Purpose | +|-------|---------| +| `clip` | CLIP model from Checkpoint Loader | +| `positive_prompt` | Positive prompt text to encode | +| `negative_prompt` | Negative prompt text to encode | +| `enhanced_prompt` | Optional enhanced/refined positive prompt | +| `edit_image_list` | Optional image list for editing mode | +| `edit_vae` | Optional VAE for image editing | + +#### Settings: + +| Setting | Purpose | +|---------|---------| +| `negative_strength` | Strength multiplier for negative guidance (default 1.20) | +| `use_int_style` | Use internal style preprocessing (ON/OFF) | +| `int_style_pos` | Positive internal style application (None, custom, etc.) | +| `int_style_pos_strength` | Strength of positive style (default 1.00) | +| `int_style_neg` | Negative internal style (None, custom, etc.) | +| `int_style_neg_strength` | Strength of negative style (default 1.00) | +| `adv_encode` | Advanced encoding mode (ON/OFF) | +| `token_normalization` | Token normalization method: mean, max, etc. | +| `weight_interpretation` | Weight interpretation: comfy++, A1111, etc. | +| `enhanced_prompt_usage` | How to use enhanced prompt: T5-XXL, append, etc. | +| `enhanced_prompt_strength` | Strength of enhanced prompt (default 1.00) | +| `opt_pos_strength` | Optimization strength for positive (default 1.00) | +| `opt_neg_strength` | Optimization strength for negative (default 1.00) | +| `style_handling` | How to handle style injection: Style-prompt merge, etc. | +| `style_position` | Where to inject style: Style to end of prompt, etc. | +| `style_swap` | Swap style to T5/L vs prompt to default clip (ON/OFF) | +| `style_pos_strength` | Style positive strength (default 1.00) | +| `style_neg_strength` | Style negative strength (default 1.00) | +| `L_strength` | L-strength modifier (default 1.00) | + +#### Outputs: + +| Output | Purpose | +|--------|---------| +| `COND+` | Positive CLIP conditioning | +| `COND-` | Negative CLIP conditioning | +| `PROMPT+` | Positive prompt text (pass-through) | +| `PROMPT-` | Negative prompt text (pass-through) | +| `T5XXL_PROMPT` | T5-XXL specific encoding | +| `PROMPT_L` | L-layer specific encoding | +| `PROMPT_L` | Additional L-layer variants (model-dependent) | +| `CONTROL_DATA` | Encoding metadata | + +--- + +### Primiere KSampler + +**Purpose:** Execute diffusion sampling process to generate latent representations from noise. + +#### Inputs: + +| Input | Purpose | +|-------|---------| +| `model` | Diffusion model from Checkpoint Loader | +| `seed` | Seed for sampling reproducibility | +| `steps` | Number of sampling steps (higher = more quality, slower) | +| `cfg` | Classifier-free guidance scale (higher = stronger prompt adherence) | +| `sampler_name` | Sampling algorithm: euler, dpmpp, etc. | +| `scheduler_name` | Noise schedule: karras, exponential, etc. | +| `positive` | Positive CLIP conditioning from Encoder | +| `negative` | Negative CLIP conditioning from Encoder | +| `latent_image` | Starting latent tensor from Noise Latent | + +#### Settings: + +| Setting | Purpose | +|---------|---------| +| `denoise` | Denoising strength: 0.0-1.0 (1.0 = full generation, 0.0 = no change) | +| `variation_extender` | Enable variation/consistency extender (default 0.00) | +| `variation_batch_step` | Batch step for variation (default 0.00) | +| `variation_level` | Variation quality level: Maximize, Normal, etc. | +| `device` | Compute device: DEFAULT (uses model device) | + +#### Outputs: + +| Output | Purpose | +|--------|---------| +| `LATENT` | Generated latent tensor | +| `CONTROL_DATA` | Sampling metadata | + +--- + +### Primiere Aesthetic Scorer + +**Purpose:** Calculate aesthetic quality score for generated images and optionally store for checkpoint/prompt metadata. + +#### Inputs: + +| Input | Purpose | +|-------|---------| +| `image` | Image tensor to score | +| `control_data` | Optional metadata tuple from workflow | + +#### Settings: + +| Setting | Purpose | +|---------|---------| +| `get_aesthetic_score` | Calculate and return score (ON/OFF) | +| `add_to_checkpoint` | Store score with checkpoint metadata (ON/OFF) | +| `add_to_saved_prompt` | Store score with saved prompt (ON/OFF) | + +#### Outputs: + +| Output | Purpose | +|--------|---------| +| `SCORE` | Aesthetic score value (0.0-10.0 or normalized) | + +**Benefit:** Pre-compute quality metrics during generation for later filtering and sorting in Visual Checkpoint Selector and Visual Style Selector modals. + +--- + +
diff --git a/Workflow/Manual/nodes/primere_prompt_save.jpg b/Workflow/Manual/nodes/primere_prompt_save.jpg index 0b9d848..ef50c54 100644 Binary files a/Workflow/Manual/nodes/primere_prompt_save.jpg and b/Workflow/Manual/nodes/primere_prompt_save.jpg differ diff --git a/Workflow/Manual/nodes/visual_checkpoint.jpg b/Workflow/Manual/nodes/visual_checkpoint.jpg index 81737d4..50cd618 100644 Binary files a/Workflow/Manual/nodes/visual_checkpoint.jpg and b/Workflow/Manual/nodes/visual_checkpoint.jpg differ diff --git a/Workflow/Manual/nodes/visual_csv.jpg b/Workflow/Manual/nodes/visual_csv.jpg index ed6d409..5039ff0 100644 Binary files a/Workflow/Manual/nodes/visual_csv.jpg and b/Workflow/Manual/nodes/visual_csv.jpg differ diff --git a/Workflow/Manual/wf_basic_prod.jpg b/Workflow/Manual/wf_basic_prod.jpg index d97d6f7..ce1b67a 100644 Binary files a/Workflow/Manual/wf_basic_prod.jpg and b/Workflow/Manual/wf_basic_prod.jpg differ diff --git a/Workflow/Manual/wf_basic_prod_plus.jpg b/Workflow/Manual/wf_basic_prod_plus.jpg deleted file mode 100644 index f71498c..0000000 Binary files a/Workflow/Manual/wf_basic_prod_plus.jpg and /dev/null differ