InvokeAI 6 Dataset Triplet Sync Node

__init__.py

from .string_splitter import SplitDatasetStringInvocation

README.md

A utility node for InvokeAI 6 designed for LoRA dataset creation. It takes a master collection of strings (formatted as WidthxHeight|Path) and breaks them into three synchronized collections: Widths (Int), Heights (Int), and Paths/Prompts (String). This allows for automated batch processing of images with varying aspect ratios without manual configuration for each shot.

1. Create a folder called split or something else in your invokeai nodes folder. Place the files in there. Make sure to restart your invoke.
2. Create a collection of strings. Enter your values such as 512x512|D:\Documents\myprompts.txt
3. Connect to iterator
4. Connect to Split Dataset String
5. Connect the output of width/height to where you need it. Connect File Path to Prompts from file

string_splitter.py

from invokeai.app.invocations.baseinvocation import (
    BaseInvocation,
    BaseInvocationOutput,
    InvocationContext,
    invocation,
    invocation_output,
)
from invokeai.app.invocations.fields import InputField, OutputField

@invocation_output("split_dataset_string_output")
class SplitDatasetStringOutput(BaseInvocationOutput):
    """Outputs for the Dataset String Splitter"""
    width: int = OutputField(description="The extracted width")
    height: int = OutputField(description="The extracted height")
    file_path: str = OutputField(description="The extracted file path or prompt")

@invocation(
    "split_dataset_string",
    title="Split Dataset String",
    tags=["dataset", "string", "size"],
    category="util",
    version="1.0.2",
)
class SplitDatasetStringInvocation(BaseInvocation):
    """Splits a string like '832x1216|C:/path' into separate sockets"""
    input_string: str = InputField(default="832x1216|C:/path/to/file.txt", description="Format: WidthxHeight|Path")

    def invoke(self, context: InvocationContext) -> SplitDatasetStringOutput:
        try:
            # 1. Split the path from the resolution
            parts = self.input_string.split('|')
            res_part = parts[0].strip()
            file_path = parts[1].strip() if len(parts) > 1 else ""
            
            # 2. Split width and height
            dimensions = res_part.lower().split('x')
            width = int(dimensions[0].strip())
            height = int(dimensions[1].strip())

            return SplitDatasetStringOutput(
                width=width,
                height=height,
                file_path=file_path
            )
        except Exception:
            # Fallback if the string is mangled to prevent workflow crash
            return SplitDatasetStringOutput(width=512, height=512, file_path=self.input_string)