This extension provides various utility nodes. Inputs(prompt, styles, dynamic, merger, ...), Outputs(style pile), Dashboard(selectors, loader, switch, ...), Networks(LORA, Embedding, Hypernetwork), Visuals(visual selectors, )
Git link: https://github.com/CosmicLaca/ComfyUI_Primere_Nodes
<a href="./Workflow/readme_images/latest_workflow.png" target="_blank"><img src="./Workflow/readme_images/latest_workflow.jpg" width="400px"></a>
<hr>Try load 'Primere_full_workflow.json' from the 'Workflow' folder, specially after git pull if the previous workflow failed because nodes changed by development. This workflow contains most of fresh developed nodes, but some 3rd party nodes and models used
1; Install missing Python libraries if nodepack not start for first try. Activate Comfy venv and use 'pip install -r requirements.txt' at the root folder of Primere nodes (or check error messages and install missing Python libs manually).
2; If node-pack started, load Primere_minimal_workflow and Primere_basic_workflow from the 'Workflow' folder for first test. All nodes visible under the 'Primere Nodes' submenu if you need for custom workflow. If some other nodes missing and red in loaded workflow, download or delete unloaded 3rd party nodes.
3; The Primere_full_workflow.json is the most complex workflow, using most of developed nodes. But the Primere_minimal_workflow.json is the simplest workflow with less required nodes. If the complex latest workflow not start or failed, please test out the basic or minimal instead. If you save own workflow with older developed nodes, try 'Fix node (recreate)' menu on right-click after git pull.
4; Rename 'styles.example.csv' on the 'stylecsv' folder to 'syles.csv' if you want to use 'Primere Styles' node with your own prompts. If you keep or rename the original 'styles.example.csv', you will see image previews for included example prompts.
5; Sometime the node development changes existing nodes, so the previous workflows may failed after git pull, usually by invalid/deprecated/missing input values. Then use right-click + 'Fix node (recreate)' menu and maybe need to rewire changed nodes, or load the attached example workflows again if updated from Workflow folder.
6; Maybe you have missing wildcard files (https://civitai.com/tag/wildcard), and sometime the wildcard decoder sending error if source file not found. If you have your own or downloaded custom wildcard files, just copy/symlink all to the 'wildcards' folder.
7; Don't overwrite attached example workflows, because the git pull will write back them to the original. If you modify them, save as... to another name and path.
<hr><a href="./Workflow/readme_details/WFComparison.md" target="_blank"><img src="./Workflow/readme_images/WFComparison.png" height="150px"></a>
<hr>Visual submenu contains similar functions like within Inputs and Networks submenu, but the selection (for example checkpoints, loras, lycoris, embeddings, styles from style.csv and hypernetworks) possible by image previews on modal. Very similar than in several themes of A1111, but you must create/save previews to right path. Create and save images as previews to the right path and name, details later. Supported previews formats: .jpg, .preview.jpg, .jpeg, .preview.jpeg, .png, .preview.png. Don't use downloaded generated large files because the long modal loading time. The preview height in visual selector is only 220px, so don't use upscaled or original/downloaded images as preview. Dowscale your previews height to max 250-300 px, and set .jpg image quality to ~50% for faster loading. ACDSee do it automatically at Tools->Batch->Resize menu if you already have large images. Semi-automatic preview saver node available, read details later. Checkpoint and additional networks files have a badge with SD or SDXL version. The version info is cached, only one time needed to read and store, so the very first loading time little longer. When you use your checkpoint or networks first time, the version info will be saved to the 'Nodes\.cache\.cache.json' file, next time will be read back. About automatic mass version caching read more later.
If you need version info of all your files for visual modal badges, you can use terminal helper files from the 'terminal_helpers' subdir:
Unfortunately the result is not perfect :(. You must check and maybe modify the version labels on your models and network files. If the result failed or unknown, you can modify and correct the .cache.json manually. Git pull will keep your edited cache file. The embedding cache helper can't read the right version of embedding files, after first run all files will be marked to SD version. You must modify and replace failed SD embeddings to SDXL in the .cache.json manually.
<hr> Example of visual checkpoint selector if preview available:<a href="./Workflow/readme_images/pvisualmodal.jpg" target="_blank"><img src="./Workflow/readme_images/pvisualmodal.jpg" height="340px"></a>
<hr><a href="./Workflow/readme_images/pimgsaveas.jpg" target="_blank"><img src="./Workflow/readme_images/pimgsaveas.jpg" height="420px"></a>
Visual selector for checkpoints. You must mirror (replicate) your original checkpoint subdirs (not the checkpoint files!) to ComfyUI\web\extensions\PrimerePreviews\images\checkpoints\ path but only the preview images needed, same name as the checkpoint. Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image. As extra features you can enable/disable modal with 'show_modal' switch, and exclude files and folders from modal starts with . (point) character if show_hidden switch is off.
<a href="./Workflow/readme_images/pvmodal.jpg" target="_blank"><img src="./Workflow/readme_images/pvmodal.jpg" height="120px"></a>
<hr>Same as than the 'Primere LORA' node, but with preview images of selection modal.
You must mirror your original lora subdirs (not your lora files!) to ComfyUI\web\extensions\PrimerePreviews\images\loras\ folder but only the preview images needed, same name as the lora files.
Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image.
As extra features you can enable/disable modal with 'show_modal' switch, and exclude files and folders from modal starts with . (point) character if show_hidden switch is off.
<a href="./Workflow/readme_images/pvlora.jpg" target="_blank"><img src="./Workflow/readme_images/pvlora.jpg" height="300px"></a>
<hr>Same as than the 'Primere LYCORIS' node, but with preview images of selection modal.
You must mirror your original lycoris subdirs (not your lycoris files!) to ComfyUI\web\extensions\PrimerePreviews\images\lycoris\ path but only the preview images needed, same name as the lyco files.
Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image.
As extra features you can enable/disable modal with 'show_modal' switch, and exclude files and folders from modal starts with . (point) character if show_hidden switch is off.
<a href="./Workflow/readme_images/pvlyco.jpg" target="_blank"><img src="./Workflow/readme_images/pvlyco.jpg" height="200px"></a>
<hr>Same as than the 'Primere Embedding' node, but with preview images of selection modal.
You must copy your original embedding subdirs (not your embedding files!) to ComfyUI\web\extensions\PrimerePreviews\images\embeddings\ path but only the preview images needed, same name as the embedding file.
Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image.
As extra features you can enable/disable modal with 'show_modal' switch, and exclude files and folders from modal starts with . (point) character if show_hidden switch is off.
<a href="./Workflow/readme_images/pvembedd.jpg" target="_blank"><img src="./Workflow/readme_images/pvembedd.jpg" height="300px"></a>
<hr>Same as than the 'Primere Hypernetwork' node, but with preview images of selection modal.
You must copy your original hypernetwork subdirs (not your hypernetwork files!) to ComfyUI\web\extensions\PrimerePreviews\images\hypernetworks\ path but only the preview images needed, same name as the hypernetwork file.
Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image.
If you have hypernetwork files from unknown source, set 'safe_load' switch to true. With this settings sometime your hypernetwork settings will be ignored, but your computer stay safe.
As extra features you can enable/disable modal with 'show_modal' switch, and exclude files and folders from modal starts with . (point) character if show_hidden switch is off.
<a href="./Workflow/readme_images/pvhyper.jpg" target="_blank"><img src="./Workflow/readme_images/pvhyper.jpg" height="200px"></a>
<hr>Same as than the 'Primere Styles' node, but with preview images of selection modal.
You must create .jpg images as preview with same name as the style name in the list, but space characters must be changed to _. For example if your style in the list is 'Architecture Exterior', you must save Architecture_Exterior.jpg to the path: ComfyUI\web\extensions\PrimerePreviews\images\styles
Much easier if you use Primere Image Preview and Save as... node for automatic preview creation from your generated image.
The styles.example.csv included, if you rename to styles.csv you will see example previews, and you can insert your own custom prompts to styles.csv.
<a href="./Workflow/readme_images/pvstyles.jpg" target="_blank"><img src="./Workflow/readme_images/pvstyles.jpg" height="300px"></a>
<hr>Under this submenu you can found nodes for detailer/refiner nodes, and required one more Primere Refiner Prompt node from the Inputs menu. For these nodes you have to download ultralitics bbox and segmentation models from here: https://huggingface.co/Bingsu/adetailer/tree/main or use Comfy's internal model downloader (this is much easier). Have to save these models to ComfyUI\models\ultralytics\segm\ and ComfyUI\models\ultralytics\bbox\ paths, the Comfy model manager save these models to right path automatically. For the included workflow these models required, but maybe you don't need all models if created your own custom workflow.
<a href="./Workflow/readme_images/pdetsmallfaces.jpg" target="_blank"><img src="./Workflow/readme_images/pdetsmallfaces.jpg" height="210px"></a>
<a href="./Workflow/readme_images/pdetlargefaces.jpg" target="_blank"><img src="./Workflow/readme_images/pdetlargefaces.jpg" height="270px"></a>
<hr>This node select segs and bbox model, but for three models: yolov8s, deepfashion2_yolov8s and facial_features_yolo8x you must select keyword label too. When you use one of these models with right label selection, the segmentation result will follow your selected label. Another models no need label. You can use these nodes for workflow result by new prompts, but you can use if the input is existing image only. Load/test attached civitai-[what]-refiner.json workflows how to use these nodes if you want to refine your existing images. You can On/Off this node anytime by switch and triggers, and you can play with available parameters.
Two trigger input available on this node, trigger_high_off and trigger_low_off. These input fields are numerical inputs, mean the percentage of original image area. For example 10 in trigger mean, the segmented area is the 10 percent of original image. Both are designed to automatically switch on/off the node by the area of segmented image. The good trigger value depending on the segmented area compare to the source image. If you want to ignore segments smaller than 5% of input image, add 5 to trigger_low_off input, and segments under 5% of original pixels will be ignored. This is useful if the segment (for example mouth) too small to do correct refining. The trigger_high_off switch off the node if the segmented area higher percent than this field value. For example if the face is always good if larger than 10 percent of original image area, enter 10 to the trigger_high_off input, and the node will process segments only if the segmented area less than 10% of original. In the example workflow for face detailers I using trigger_high_off = 1, because if the area of segmented face less than 1%, then I need the node for fix small faces. If larger than 1%, no need fixes because usually good enough. The right value depending on used model, prompt, and additional networks like Loras or controlnet settings. For mouth I using trigger_low_off = 0.25, because if the area of mouth less than 0.25%, no need to repair, only if larger. For hand fixer I set trigger_high_off to 5, because if the hand's area is larger than 5%, usually no need to fix/detail. All settings depending in workflow settings and the input image.
<a href="./Workflow/readme_images/pimgsegments.jpg" target="_blank"><img src="./Workflow/readme_images/pimgsegments.jpg" height="350px"></a>
This node create detailed/refined output by input image and segments. Node must be used together with Image Segments and Refiner Prompt. The output of this node can be upscaled or saved, maybe connected to the next refiner. Detailer group example included to the Primere_full_workflow.json you can check it for your own ideas and settings, or test only detailers in attached civitai-[what]-refiner.json files.
<a href="./Workflow/readme_images/panydetailer.jpg" target="_blank"><img src="./Workflow/readme_images/panydetailer.jpg" height="380px"></a>
The node using aesthetic scorer to measure the quality of the detailed segment. If the use_aesthetic_scorer on (ignore_if_worse) the result will be ignored if the aesthetic score is lower than the original segment's score. The a-scorer model sometime failed, this is why the feature can be off (always_refine) so you can ignore this feature. If the score metered (ignore_if_worse), and the detailed segment's score is better than than the original, green badge show in the top-right corner, that the refined segment accepted. If the a-score of result lower than the original, red badge show in the corner, and the segmented part of image not changed to refined.
<a href="./Workflow/readme_images/padetailer_example.jpg" target="_blank"><img src="./Workflow/readme_images/padetailer_example.jpg" height="300px"></a>
<hr>2 input fields within one node for positive and negative prompts. 3 additional fields appear under the text inputs:
If you set these fields, (where 'None' mean not set and use dashboard settings) the workflow will use all of these settings for rendering your prompt instead of settings in 'Dashboard' group.
<a href="./Workflow/readme_images/pprompt.jpg" target="_blank"><img src="./Workflow/readme_images/pprompt.jpg" height="180px"></a>
<hr>Style (.csv) file reader, compatible with A1111 syles.csv, but little more than the original concept. The file must be copied/symlinked to the 'stylecsv' folder. Rename included 'style.example.csv' to 'style.csv' for first working example, and later edit this file manually.
<a href="./Workflow/readme_images/pstyles.jpg" target="_blank"><img src="./Workflow/readme_images/pstyles.jpg" height="160px"></a>
<hr>Prompts and additional data must be stored in the .toml file. This node dynamically read and organize the custom file content. The example file with data schema is on the path Toml/prompts.example.toml, just rename to the prompts.toml end edit/include your own prompt.
Follow the file schema for your own prompts but don't forget to rename the attached example file to prompts.toml.
<a href="./Workflow/readme_images/ppromptorganizer.jpg" target="_blank"><img src="./Workflow/readme_images/ppromptorganizer.jpg" height="280px"></a>
<hr><a href="./Workflow/readme_images/pdynamic.jpg" target="_blank"><img src="./Workflow/readme_images/pdynamic.jpg" height="80px"></a>
<hr><a href="./Workflow/readme_images/pimgrecycler.jpg" target="_blank"><img src="./Workflow/readme_images/pimgrecycler.jpg" height="340px"></a>
<a href="./Workflow/readme_images/pmetadistribitions.jpg" target="_blank"><img src="./Workflow/readme_images/pmetadistribitions.jpg" height="320px"></a>
<hr>This node convert A1111 embeddings to Comfy embeddings. Use after dynamically decoded prompts (booth text and style). No need to modify manually styles.csv from A1111 if you use this node.
<a href="./Workflow/readme_images/pemhandler.jpg" target="_blank"><img src="./Workflow/readme_images/pemhandler.jpg" height="80px"></a>
<hr>This node merge two different Lora stacks, SD and SDXL. The output is useful to store Lora settings to the image meta.
<hr>With Lora stackers you can read model keywords. This node merge all selected Lora keywords to one string, and send to prompt encoder.
<hr>This node merge two different Lycoris stacks, SD and SDXL. The output is useful to store Lycoris settings to the image meta.
<hr>With Lycoris stackers you can read model keywords. This node merge all selected Lycoris keywords to one string, and send to prompt encoder.
<hr>This node merge positive and negative SD and SDXL embedding tags, to send them to the prompt encoder.
<hr>This node collect (and merge) settings of Lora, Lycoris, Embedding, Hypernetwork nodes for metada saving.
<a href="./Workflow/readme_images/pndatacollector.jpg" target="_blank"><img src="./Workflow/readme_images/pndatacollector.jpg" height="120px"></a>
<hr>This node collect (and merge) all required data for image saver. Important to use if you want enjoy all node functions. Check attached full workflow how to use.
<a href="./Workflow/readme_images/pmetatuplecollector.jpg" target="_blank"><img src="./Workflow/readme_images/pmetatuplecollector.jpg" height="80px"></a>
<hr>This node select between SD and SDXL VAE if model_version input is correct.
<a href="./Workflow/readme_images/pvaeselector.jpg" target="_blank"><img src="./Workflow/readme_images/pvaeselector.jpg" height="80px"></a>
<hr>Another dual prompt input, but specially for refiners and detailers. You can connect original prompts too to this node, and set the weights of all inputs. Text and cond outputs are available.
<a href="./Workflow/readme_images/prefprompt.jpg" target="_blank"><img src="./Workflow/readme_images/prefprompt.jpg" height="500px"></a>
<hr>Use this separated node for sampler/meta reader inputs. If you use LCM and Turbo modes, you need 3 with different settings of this node. See and test the attached example workflow. This node the merged version of previous two: 'Primere Sampler Selector' and 'Primere Steps & Cfg'
<a href="./Workflow/readme_images/psamcfgsel.jpg" target="_blank"><img src="./Workflow/readme_images/psamcfgsel.jpg" height="140px"></a>
<hr>Use this node to switch between Normal, LCM, Cascade, Lightning, Playground, Hyper-SD, Flux and Turbo modes in whole rendering process. Use several sampler and cfg/steps settings to the inputs (one of them must be compatible with LCM settings, another must follow Turbo, Lightning, Playground and Cascade rules), and connect this node output to the sampler/exif reader, like in the example workflow. The 'MODEL_CONCEPT' output important for CKPT loader, Image refiners, and the Exif reader for correct rendering. Hyper-SD and Flux UNET files must be saved to Comfy's unet path, not as checkpoint!
<a href="./Workflow/readme_images/pmodelconcept.jpg" target="_blank"><img src="./Workflow/readme_images/pmodelconcept.jpg" height="500px"></a>
<hr>This node is a simple VAE file selector. Use 2 nodes in workflow, 1 for SD, 1 for SDXL compatible VAE for automatized selection.
<hr>Simple checkpoint selector, but with extras:
<a href="./Workflow/readme_images/pckptselect.jpg" target="_blank"><img src="./Workflow/readme_images/pckptselect.jpg" height="80px"></a>
<hr>Use this node to convert VAE name to VAE.
<hr>Use this node to convert checkpoint name to 'MODEL', 'CLIP' and 'VAE'. Use 'model_concept' input for detect LCM and Turbo modes, see the example workflow. If you have downloaded .yaml file, and copied to the checkpoint's directory with same filename, set 'use_yaml' to true, and the loader will use read and use the config file. No need to switch off if .yaml file missing. If you find some problem or error, simply set it to false. Play with 'strength_lcm_model' and 'strength_lcm_clip' values if set LCM mode on whole workflow. This node have optional inputs if checkpoint already loaded by previous process. If 'loaded_clip', 'loaded_vae' and 'loaded_model' connected, this node will use these inputs instead of loading checkpoint again.
<a href="./Workflow/readme_images/pckpt.jpg" target="_blank"><img src="./Workflow/readme_images/pckpt.jpg" height="150px"></a>
<hr>Use this node if you have more than one prompt input (for example several half-ready test or development prompts). Connect prompts/styles node outputs to this node inputs and set the right index at the bottom. To connect 'subpath', 'model', and 'orientation' inputs are optional, only the positive and negative prompt required.
Very important: don't remove the connected node from the middle or from the top of inputs. Connect nodes in right queue, and disconnect them only from the last to first. If you getting js error because disconnected inputs in wrong queue, just reload your browser and use 'reload node' menu with right click on node.
<a href="./Workflow/readme_images/prpmptswitch.jpg" target="_blank"><img src="./Workflow/readme_images/prpmptswitch.jpg" height="150px"></a>
<hr>Use only one seed input for all. A1111 look node, connect this one node to all other seed inputs.
<hr>This node generate 'empty' latent image, but with several noise settings, what control the final images. You can randomize these setting between min. and max. values using switches, this cause small difference between generated images for same seed and settings, but you can freeze your noise and latent image if you disable variations of random noise generation.
<a href="./Workflow/readme_images/platent.jpg" target="_blank"><img src="./Workflow/readme_images/platent.jpg" height="280px"></a>
<hr><a href="./Workflow/readme_images/pencoder.jpg" target="_blank"><img src="./Workflow/readme_images/pencoder.jpg" height="320px"></a>
<hr><a href="./Workflow/readme_images/pres.jpg" target="_blank"><img src="./Workflow/readme_images/pres.jpg" height="180px"></a>
<hr>Multiply the base image size for upscaling. This node upscale original images to the value of megapixels (upscale_to_mpx). No need to use model concepts or versions, just the original image size needed. Just switch off 'use_multiplier' on this node if you don't need to resize the original image.
If your upscaler failed because low memory error, try to switch on 'triggered_prescale' and set right values to the input fields under this switch. This function resize the source image before upscaling to right size:
<a href="./Workflow/readme_images/presmulmpx.jpg" target="_blank"><img src="./Workflow/readme_images/presmulmpx.jpg" height="180px"></a>
<hr>This node remove Lora, Lycoris, Hypernetwork and Embedding (booth A1111 and Comfy) from the prompt and style inputs. Use switches what network(s) you want to remove or keep in the final prompt. Use 'remove_only_if_sdxl' if you want keep all of these networks for SD1.5 models, and remove only if SDXL checkpoint selected. Important notice: for loras, lycoris and hypernetworks you don't need original tags in the prompt (for example: <lora:your_lora_name>). If you keep original lora and hypernetwork tags in the prompt you cant sure your image result use the lora only, or use the tag string only (or booth) in the prompt. I recommend always to remove lora and hypernetwork tags, but you can try what happen if keep. You must remove original tags after 'Primere Network Tag Loader', because after prompt cleaner no tags available for tag loader. The example workflow using 2 of this nodes, one for SD, one for SDXL workflow.
<a href="./Workflow/readme_images/ppcleaner.jpg" target="_blank"><img src="./Workflow/readme_images/ppcleaner.jpg" height="120px"></a>
<hr>This node loads additional networks (Lora, Lycoris and Hypernetwork) to the CLIP and MODEL. You can read and use Lora (lora:[your model name]), Lycoris (lyco:[your model name]) and Hypernetwork (hypernetwork:[your model name]) keywords to send to prompt encoder or the keyword merger like in the example workflow. Hypernetwork is harmful, because can run any code on your computer, so set 'process_hypernetwork' to False on this node or download them from reliable source only If you have hypernetwork files from unknown source, set 'safe_load' switch to true. With this settings sometime your hypernetwork tags will be ignored, but your computer stay safe.
<a href="./Workflow/readme_images/pnettagload.jpg" target="_blank"><img src="./Workflow/readme_images/pnettagload.jpg" height="280px"></a>
<hr>This node loads model keyword. You can read and use model keywords to send directly to prompt encoder like in the example workflow. The idea based on A1111 plugin, but something different.
<a href="./Workflow/readme_images/pmodkeyw.jpg" target="_blank"><img src="./Workflow/readme_images/pmodkeyw.jpg" height="200px"></a>
<hr>This node is image preview, but with save as feature.
<a href="./Workflow/readme_images/pimgsaveas2.jpg" target="_blank"><img src="./Workflow/readme_images/pimgsaveas2.jpg" height="300px"></a>
<hr>This node save the image, but with/without metadata, and save meta to .json/.txt file if you want. Get metadata from the Exif reader node only, and use optional 'preferred_subpath' input if you want to overwrite the node settings by several prompt input nodes. Set 'output_path' input correctly, depending your system.
<a href="./Workflow/readme_images/pimgsaver.jpg" target="_blank"><img src="./Workflow/readme_images/pimgsaver.jpg" height="320px"></a>
<hr>Use this node to display 'any' output values of several nodes like prompts or metadata (metadata is formatted). See the example workflow for details.
<hr>Use this node to display simple text (not tuples or dict).
<hr>Use this node in the workflow if you don't need Primere Meta Reader node. This node collect required metadata for Primere Meta Saver, the data will be stored to .jpg exif or .png pnginfo, then you can read back and recycle your previous prompts and settings by Primere Meta Reader. Check 'Primere_advanced_workflow.json' how to use this node.
<a href="./Workflow/readme_images/pmetacoll.jpg" target="_blank"><img src="./Workflow/readme_images/pmetacoll.jpg" height="250px"></a>
<hr>Get the aesthetic score of your generated image.
<a href="./Workflow/readme_images/pascorer.jpg" target="_blank"><img src="./Workflow/readme_images/pascorer.jpg" height="180px"></a>
<hr>Sampler using the 'model_concept' input this node automatically handle Turbo, Flux and Cascade modes, no need another workflow or extra node. You can select device (CPU or GPU), and use 'variation_extender' input for images with very small (adjustable) difference from previous one (if seed and other details freezed). This settings can be used in queued workflow.
<a href="./Workflow/readme_images/pksampler.jpg" target="_blank"><img src="./Workflow/readme_images/pksampler.jpg" height="280px"></a>
<hr>Style collection for generated images. Set and connect this node to the 'Prompt Encoder'. No forget to set and play with style strength. The source of this node is external file at 'Toml/stylepile.toml', what you can edit if you need changes.
<a href="./Workflow/readme_images/pstylepile.jpg" target="_blank"><img src="./Workflow/readme_images/pstylepile.jpg" height="200px"></a>
<hr>Style collection from Midjourney. You can attach art-style prompt to your original prompt, and get your result in several artistic style.
<a href="./Workflow/readme_images/pmidjourney.jpg" target="_blank"><img src="./Workflow/readme_images/pmidjourney.jpg" height="340px"></a>
Example images about the result of this style node. The left-top image is the original without any style, all others styled by this one, using SD1.5 model, same seed, same prompt:
<a href="./Workflow/readme_images/mjmontage.jpg" target="_blank"><img src="./Workflow/readme_images/mjmontage.jpg" height="300px"></a>
<hr>Style collection of emotions. You can attach emotion-style to your original prompt, and get your result in several emotion style. The source content will be update frequently to more emotions.
<a href="./Workflow/readme_images/pstyleemo.jpg" target="_blank"><img src="./Workflow/readme_images/pstyleemo.jpg" height="280px"></a>
Example images about the result of this style node. The left-top image is the original without any style, all others styled by this one, using SDXL model, same seed, same prompt:
<a href="./Workflow/readme_images/emomontage.jpg" target="_blank"><img src="./Workflow/readme_images/emomontage.jpg" height="300px"></a>
<hr>Lora stack for 6 loras. Important to use 'stack_version' list. Here you can select how you want to load selected Lora-s, for SD models only, for SDXL models only or for booth (Any) what not recommended. Use 2 separated Lora stacks for SD/SDXL checkpoints, and wire 'model_version' input for correct use.
<a href="./Workflow/readme_images/plora.jpg" target="_blank"><img src="./Workflow/readme_images/plora.jpg" height="260px"></a>
<hr>Lycoris files have dedicated node, working similar than the LORA stack. See example workflow, or use as LORA. If you already have downloaded LyCORIS files, you must symlink or copy to the path **ComfyUI\models\lycoris**. I recommend symlink the original source instead of copying.
<a href="./Workflow/readme_images/plyco.jpg" target="_blank"><img src="./Workflow/readme_images/plyco.jpg" height="260px"></a>
<hr>Select textual inversion called Embedding for your prompt. You have to use 2 several versions of this one, one for SD, and another one for SDXL checkpoints. Important to use 'model_version' input and 'stack_version' list, working similar than in the Lora stack. You can choose embedding placement in the prompt.
<a href="./Workflow/readme_images/pembed.jpg" target="_blank"><img src="./Workflow/readme_images/pembed.jpg" height="260px"></a>
<hr>Use hypernetwork if you already have by this node. Hypernetwork is harmful, because can run any code on your computer, so ignore/delete this node or download them from reliable source only If you have hypernetwork files from unknown source, set 'safe_load' switch to true. With this settings sometime your hypernetwork settings will be ignored, but your computer stay safe. Hypernetworks don't need seperated SD and SDXL sources, use only one stack for all, and set 'stack_version' to 'Any'.
<a href="./Workflow/readme_images/phyper.jpg" target="_blank"><img src="./Workflow/readme_images/phyper.jpg" height="220px"></a>
<hr>Use Primere Resolution MPX instead
Multiply the base image size for upscaling. Important to use 'model_version' and 'model_concept' if you want to use several multipliers for Turbo, SD and SDXL models. Just switch off 'use_multiplier' on this node if you don't need to resize the original image.
If your upscaler failed because low memory error, try to switch on 'triggered_prescale' and set right values to the input fields under this switch. This function resize the source image before upscaling to right size:
<a href="./Workflow/readme_images/presmul.jpg" target="_blank"><img src="./Workflow/readme_images/presmul.jpg" height="200px"></a>
<hr>Use Primere Model Concept Selector instead
Use this node to switch on/off LCM mode in whole rendering process. Wire two sampler and cfg/steps settings to the inputs (one of them must be compatible with LCM settings), and connect this node output to the sampler/exif reader, like in the example workflow. The 'IS_LCM' output important for CKPT loader and the Exif reader for correct rendering.
<a href="./Workflow/readme_images/plcm.jpg" target="_blank"><img src="./Workflow/readme_images/plcm.jpg" height="150px"></a>
<hr>Use Primere Samplers & Steps & Cfg instead
Select sampler and scheduler in separated node, and wire outputs to the sampler (through exif reader input in the example workflow). This is useful to separate from other nodes, and for LCM and Turbo modes you need three several sampler settings. (see the example workflow, and try to understand LCM and Turbo setting)
<a href="./Workflow/readme_images/psampler.jpg" target="_blank"><img src="./Workflow/readme_images/psampler.jpg" height="80px"></a>
<hr>Use Primere Samplers & Steps & Cfg instead
Use this separated node for sampler/meta reader inputs. If you use LCM and Turbo modes, you need 3 with several settings of this node. See and test the attached example workflow.
<a href="./Workflow/readme_images/psteps.jpg" target="_blank"><img src="./Workflow/readme_images/psteps.jpg" height="80px"></a>
<hr>Use Primere image recycler instead
This node output sending lot of data to the workflow from exif/meta or pnginfo if it's included to selected image, like model name, vae and sampler name or settings. Use this node to distribute settings, and simple off the 'use_exif' switch if you don't want to render image by this node, then you can use your own prompts and dashboard settings instead.
Use several settings of switches what exif/meta data you want/don't want to use for new image rendering. If switch off something, dashboard settings (this is why must be connected this node input) will be used instead of image included exif/meta.
<a href="./Workflow/readme_images/pexif.jpg" target="_blank"><img src="./Workflow/readme_images/pexif.jpg" height="300px"></a>
<hr>