Comfy-UmiAI

CameraAngleSelector

Select camera angles using a 3D visual interface rendered directly in the node

Camera · 0 inputs · 1 parameter · 1 output

UmiAIWildcardNode

Seeded wildcard processor for repeatable prompt variation.

UmiAI · 9 outputs

UmiAIWildcardNodeLite

Alias so workflows saved with the old 'UmiAIWildcardNodeLite' class name keep loading. DEPRECATED hides it from the node menu on newer frontends.

UmiAI · 13 outputs

UmiBypassModelSwitch

Chooses between basemodel and referencemodel using the same matchedlist and matchindex pattern. This is useful for workflows where a prompt phrase should switch to a reference or edit model.

UmiAI · 3 inputs · 1 parameter · 1 output

UmiCharacterCreator

Create or update character profiles with VNCCS-compatible structure. Generates folder structure, config, and outputs for downstream nodes.

UmiAI/Character · 7 outputs

UmiDatasetGenerator

Create LoRA training dataset from character images. Copies images to a lora folder and generates caption files.

UmiAI/Character · 0 inputs · 5 parameters · 2 outputs

UmiEmotionGenerator

Generate emotion prompts from character YAML data. Reads the Emotions field from character profile and outputs prompts for each selected emotion.

UmiAI/Character · 0 inputs · 2 parameters · 2 outputs

UmiEmotionStudio

Enhanced Emotion Generator with visual emotion picker. Uses emotions.json config for categorized emotions with descriptions. Supports SDXL and QWEN prompt styles.

UmiAI/Character · 3 outputs

UmiModelManager

Passes the supplied repository ID through as a string.

UmiAI/Manager · 0 inputs · 1 parameter · 1 output

UmiModelSelector

Looks up a named model and version in a repository's model_updater.json and returns its normalized local model path; returns an empty string when the model cannot be resolved.

UmiAI/Manager · 0 inputs · 1 parameter

UmiPoseGenerator

Pose Generator for creating OpenPose images from JSON pose data. Creates a 6x2 grid of 12 poses in OpenPose format for ControlNet. Compatible with VNCCS 3D pose editor.

UmiAI/Pose · 0 inputs · 3 parameters · 1 output

UmiPositionControl

Generate camera position prompts for QWEN-style angle control. Uses azimuth/elevation/distance to create position prompts.

UmiAI/Camera · 0 inputs · 4 parameters · 1 output

UmiPromptInspector

Compares an input prompt and a processed prompt. It returns: - diffjson: added and removed prompt tokens. - lintwarnings: profile and LoRA-related warnings. - sectionsjson: parsed prompt sections.

UmiAI/Prompt · 3 outputs

UmiPromptPreset

Applies one preset from promptpresets.yaml to a prompt and negative prompt. Placement can append, prepend, or replace the positive prompt. Preset negatives are appended to the negative prompt.

UmiAI/Prompt · 3 outputs

UmiPromptProfile

Applies a named profile from promptprofiles.yaml. Profiles can add positive defaults, merge negative defaults, and emit warnings when prompts look too short or mismatched for the target model family. Core profiles include None, Illustrious, Pony, SDXL, and Flux. The Anima/edit overlay adds Anima Base and its Anima presets at runtime.

UmiAI/Prompt · 4 outputs

UmiPromptSyntaxLint

Checks prompt syntax without expanding randomness. It reports errors, warnings, and a structured JSON report for: - unclosed delimiters, - wildcard range issues, - malformed LoRA options, - missing wildcard and prompt files, with typo suggestions and fallback notes, - missing presets, - profile warnings, - suspicious prompt joins, - negative syntax counts.

UmiAI/Prompt · 3 outputs

UmiSaveImage

Save images with Umi prompt metadata embedded for the Image Browser.

UmiAI · 5 inputs · 1 parameter · 0 outputs

UmiSpriteGenerator

Collect character sprites from the character folder structure. Scans character's Sheets folder for all costume/emotion combinations.

UmiAI/Character · 0 inputs · 1 parameter · 2 outputs

UmiTextBypass

Create a small ComfyUI node that: 1) Receives a boolean signal from the UmiAI Wildcard Processor 2) Sits inline between any data stream (e.g., IMAGE) 3) Automatically bypasses downstream nodes based on the boolean signal Example: If the wildcard processor's text contains simple background, it outputs bypassmatched=True. The bypass node receives this and allows downstream nodes to run. Otherwise, they're skipped. How it works: - UmiAI Wildcard Processor has a bypassphrase input (optional) - It checks if the generated prompt contains that phrase - Outputs a bypassmatched boolean - UmiTextBypass receives this boolean as input - During execution, a custom hook skips nodes connected to UmiTextBypass when matched=False - This happens dynamically DURING the same execution run! - Added bypassphrase optional input (STRING) - Added bypassmatched output (BOOLEAN) - Checks if generated prompt contains the bypassphrase - Outputs True if phrase found, False otherwise - Node class: UmiTextBypass in nodeslite.py - Inputs: - matched (BOOLEAN, forceInput) - connects to wildcard processor's bypassmatched - passthroughtype (dropdown): IMAGE, LATENT, CONDITIONING, MODEL, CLIP, STRING - Optional inputs for each type (image, latent, conditioning, model, clip, string) - Outputs: - One output for each type (image, latent, conditioning, model, clip, string) - The node: 1. Receives the matched boolean from wildcard processor 2. Passes through the selected data type 3. Execution hook checks the matched value 4. If matched=False, skips all nodes connected to this bypass node's outputs - init.py regist…

UmiAI · 7 inputs · 3 parameters · 6 outputs

UmiVisualCameraControl

Visual camera control with interactive canvas widget. Receives JSON data from the camera widget JS.

UmiAI/Camera · 0 inputs · 1 parameter · 1 output