Aller au contenu
Node packVérifiéExterne
Ouvrir la source

ComfyUI-RyuuNoodles

Collection of a few custom nodes for ComfyUI made mainly for personal use

Mainteneur@DraconicDragonUnratedClasse

Profil du package

Statut
Vérifié
Éditeur
@DraconicDragon
Source
GitHub

LIREMOI

Collection of one or more custom nodes for ComfyUI made mainly for personal use ...rawr...

ComfyUI Registry: https://registry.comfy.org/nodes/RyuuNoodles

To see the settings this pack adds, please see SETTINGS.md.

Table of Contents

Features/Nodes

Live Token Counter on Textboxes

Yeeeeeee. (not shown in the GIF but it respects prompt weighting (word_a(word_b:1.2)))
Does not support ComfyUI frontend's new Vue Nodes/"Nodes 2.0"

No CLIP input required. With this repo/custom nodes you'll have the ability to add a token counter with variable tokenizers to any node/widget you desire (preferably a multiline text widget but you do you lol). This is done by going to the "RyuuNoodles 🐲" settings page and adding the internal node name and the widget name of that node you want the counter to be on top of.
BREAK is also supported (You have to enable it in settings; Also note that this does not add any functionality, it just increases the token count.) Not supported currently: embeddings, wildcard and prompt control/schedule syntax (duh)

Format

Node_Name.widget_name:Tokenizer1,Tok2,Tok3;Node Name 2.widget_name:Tok1; etc...

  • Supported tokenizers: CLIP_L, T5, T5_Fast, UMT5, Gemma2, Gemma3, LlaMA3, Qwen2.5VL, AuraFlow (case insensitive)
  • RegEx is not supported; Spaces are allowed, they will be stripped anyway (node name is excluded from stripping too naughty)
  • The tokenizer types will appear in the counter in the order you added them.

To get the internal node name:
Right-click a node and select 'Properties Panel'. The value for Node name for S&R is what you want to copy.

To get the widget name:
Usually it's simply the placeholder text (or input name on single line text widgets).
(if there's a custom placeholder text set... then uh good luck, maybe need to read the code of that custom node in that case?)

Example (also default value)

Ryuu_TokenCountTextBox.input_text:CLIP_L,T5_FAST;CLIPTextEncode.text:CLIP_L,T5_FAST;
This adds the CLIP-L and T5🚀 (🚀= Fast) token counters ontop of the RyuuNoodles Textbox and the Comfy Core Clip Text Encode (Prompt) node.
Also: The Text Multiline node from WAS node suite shown in the gif above has the node name Text Multiline

  • Wildcards/Prompt Control or Scheduling etc:
    Not supported. The counter reads what's inside the widget it's assigned to, the wildcard processor nodes only process the text when queued, same with prompt scheduling. It's possible to make a wildcard node that works, but it would have to work with an external seed/randomization thingy and whatnot, and it probably wouldnt help with reproducability/saving workflow.
    comfyui-ppm and JNodes (not KJNodes) offer a Token Counter that executes when queued.

  • When you install this node pack it will do a one time download with huggingface transformers on first start after install of a few MBs of files (most are ~1 MB json files, more or less) for the supported tokenizers. (total is probably like less than 30mb maybe? dont take my word on it)
    The tokenizers are now loaded on dema

Nœuds dans ce pack

Nœuds 18
  • Ryuu_CleanStringAdvanced
  • Ryuu_ColorMatch
  • Ryuu_ExtractAndSaveLora
  • Ryuu_FallbackPassthrough
  • Ryuu_FallbackSwitchAny
  • Ryuu_FallbackSwitchImage
  • Ryuu_FallbackSwitchLatent
  • Ryuu_FloatPlain
  • Ryuu_FloatPlainLarger
  • Ryuu_FloatSlider
  • Ryuu_IntSlider
  • Ryuu_IsMultipleOf
  • Ryuu_ScaleToMultiple
  • Ryuu_ScaleToMultipleAdvanced
  • Ryuu_ScaleToMultipleLatentSizePicker
  • Ryuu_TextEncoderDiffCheck
  • Ryuu_TokenCountTextBox
  • Ryuu_WaitNode

Ressources connexes

16 connexions
Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v4Utilisé parLTX2.3 GGUF Simple i2v + Last Frame 16GB/low vram — v3 — v5Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v8Utilisé parMiniMax H3 Simple workflow | 16GB VRAM 32GB RAM quite fast — v1 — v2Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v7Utilisé parMiniMax H3 Simple workflow | 16GB VRAM 32GB RAM quite fast — v1 — v3Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v3Utilisé parBlink's universal workflow | CivitAI metadata, upscaling, FaceDetailer (+indexing elements), inpainting, controlnet, IPAdapter (regional conditioning) and token counter (SDXL/PONYV6/ILLUSTRIOUS/NOOBAI-XL/ANIMAGINEXL4/SD1.5) ComfyUI beginner friendly all in one — v5Utilisé parBlink's ANIMA workflow - ti2i AIOUtilisé parMiniMax H3 Simple workflow | 16GB VRAM 32GB RAM quite fast — v1 — v4Utilisé parLTX2.3 GGUF Simple i2v + Last Frame 16GB/low vram — v3 — v4Utilisé parBlink's universal workflow | CivitAI metadata, upscaling, FaceDetailer (+indexing elements), inpainting, controlnet, IPAdapter (regional conditioning) and token counter (SDXL/PONYV6/ILLUSTRIOUS/NOOBAI-XL/ANIMAGINEXL4/SD1.5) ComfyUI beginner friendly all in one — v4Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v5Utilisé parKrea2 workflow | ti2i 16GB VRAM CivitAI metadata, upscaling, controlnet, FaceDetailer (+indexing elements), inpainting ComfyUI beginner friendly K2 — v2 — v6Utilisé parBlink's universal workflow | CivitAI metadata, upscaling, FaceDetailer (+indexing elements), inpainting, controlnet, IPAdapter (regional conditioning) and token counter (SDXL/PONYV6/ILLUSTRIOUS/NOOBAI-XL/ANIMAGINEXL4/SD1.5) ComfyUI beginner friendly all in one — v5

Sources

3 sources
GitHubSourceVersion c9cfb96d46aec1e916b2ea1845dde311d3173733
github.com/DraconicDragon/ComfyUI-RyuuNoodles

Vérifié 15 oct.

Extraits de sources

1 extrait

Source context: Repo DraconicDragon/ComfyUI-RyuuNoodles