# ComfyUI-CraftKit Custom nodes for ComfyUI - image resizing, dataset prep & prompt automation for LoRA training workflows. All nodes appear under the **CraftKit** category in the node menu. --- ## Nodes ### 📋 Smart Prompt Controller Cycle through up to 4 prompt lists using a single incrementing index. Counts lines automatically, selects the right prompt, and outputs which list is active - ideal for driving Switch nodes that control aspect ratio, latent size, or other per-category settings. **Typical usecases:** - LoRA dataset generation with multiple pose categories (headshots, halfbody, fullbody, tall portrait), each with their own prompt list and aspect ratio - Any batch workflow where you need to rotate through different prompt sets and switch settings per set | Input | Type | Description | |---|---|---| | Index | INT | Current position across all lists (1-indexed). Set "control after generate" to `increment` to auto-advance every run | | Prompt list 1 | STRING | First prompt list, one prompt per line | | Prompt list 2 | STRING | Second prompt list (optional) | | Prompt list 3 | STRING | Third prompt list (optional) | | Prompt list 4 | STRING | Fourth prompt list (optional) | **Outputs:** Prompt (selected line as STRING), Active list (active list number as INT → feed to Smart Profile Switch or Switch nodes) The node displays a status label showing the current position: `List 1 — 4/5 | total: 8` --- ### 🧩 Smart Profile Switch Maps an active index (1–4) to a label, width, and height. Pairs directly with Smart Prompt Controller's `active_list` output to switch resolution and filename label per category in one node — no separate Aspect Ratio + Primitive + Switch combo needed. **Typical usecases:** - Driving `EmptyLatentImage` / `EmptySD3LatentImage` width/height per prompt category - Feeding a filename suffix into `SaveImageExtended` or similar so each category's output is labeled automatically - Replacing a CR Aspect Ratio + 4× Primitive + LatentSwitch chain with a single node | Input | Type | Default | Description | |---|---|---|---| | Index | INT | 1 | Active profile (1–4). Connect to Active list from Smart Prompt Controller | | Label 1 … Label 4 | STRING | Headshots / Halfbody / Fullbody / Tall | Label for each profile, used as filename suffix | | Width 1 … Width 4 | INT | 1024 / 832 / 768 / 640 | Width for each profile | | Height 1 … Height 4 | INT | 1024 / 1216 / 1344 / 1536 | Height for each profile | **Outputs:** Label (STRING), Width (INT), Height (INT) The node displays a status label showing the active slot: `Slot 2: Halfbody | 832 × 1216` --- ### 📐 Smart Resize Resize any image (or batch) so the **longest side** equals `longest_side`, with aspect ratio preserved. Use this as a **pipeline node** — IMAGE in, IMAGE out. No files are saved to disk. **Typical usecases:** - Downscale a SeedVR2 (or other upscaler) result back to your target training size — the upscaler adds fine detail, Smart Resize brings it to the right dimensions for LoRA training - Any workflow step that needs a clean resize before the next node - Normalize mixed-resolution batches to a consistent size | Input | Type | Default | Description | |---|---|---|---| | Image | IMAGE | — | Single image or batch | | Longest side (px) | INT | 1536 | Target size for longest side. Quick presets: 512 / 768 / 1024 / 1536 — or type any custom value directly into the field | | Round to multiple of | INT | 8 | Snap dimensions to this multiple (8 = SD/Flux compatible) | | Interpolation method | ENUM | lanczos | lanczos / bicubic / bilinear / nearest | | Upscale if smaller | BOOLEAN | true | Upscale images smaller than longest_side. Turn off to only ever downscale, never upscale | **Outputs:** Image, Width, Height --- ### 📁 Smart Batch Resize Load **all images from a folder**, resize each one by longest side, and save into a subfolder. Build clean dataset filenames from a prefix, the original name, and/or a sequential counter — combined with an optional resolution suffix. Use this for **bulk preprocessing** — e.g. preparing a LoRA dataset from a folder of high-res images. It works as a **standalone node**: it needs no upstream input and no downstream connection to do its job — drop it on the canvas, point it at a folder, hit **Run Batch**, and it reads, resizes, and saves everything itself. Connecting the `images`/`count` outputs is entirely optional, e.g. for previewing results. Includes a **Browse folder** button to pick the input folder directly from the node, and quick presets (512 / 768 / 1024 / 1536) for the longest side — or type any custom value directly into the field. | Input | Type | Default | Description | |---|---|---|---| | Input folder | STRING | — | Source folder path (use Browse button or paste manually) | | Longest side (px) | INT | 1024 | Target size for longest side. Quick presets: 512 / 768 / 1024 / 1536 — or type any custom value directly into the field | | Round to multiple of | INT | 8 | Snap dimensions to this multiple | | Interpolation method | ENUM | lanczos | lanczos / bicubic / bilinear / nearest | | Filename prefix | STRING | — | Label prepended to filename — e.g. `headshot` → `headshot_photo_001_1024.jpg` | | Keep original filename | BOOLEAN | true | Include the original filename in the output name | | Add counter | BOOLEAN | false | Add a sequential 3-digit counter to each filename (001, 002, ...) | | Counter start | INT | 1 | Starting number for the counter | | Add resolution to filename | BOOLEAN | true | Append resolution to filename — e.g. `photo_1024.png` | | Create resolution subfolder | BOOLEAN | false | Append resolution to subfolder name — e.g. `resized_1024` | | Custom output subfolder | STRING | resized | Subfolder name inside the input folder | | Skip existing files | BOOLEAN | true | Skip files that already exist in the output folder | | Filename delimiter | STRING | _ | Separator between filename parts | **Outputs:** Images (list), Count --- ### 📏 Smart Resolution Multiplier Multiply image dimensions by a factor and output the results as integers. SeedVR2 takes a single `resolution` INT (the longest side) — not a width and height separately. Standard math nodes give you a FLOAT or require multiple steps to get there. This node does it cleanly in one step: give it your image and a multiplier, and it outputs `width`, `height`, and `resolution` (longest side) ready to connect directly to SeedVR2's `resolution` input. Also useful for LoRA dataset prep — using mixed resolutions in your training set generally produces better results than training on a single fixed size. Smart Resolution Multiplier makes it easy to dynamically calculate the right target size per image rather than hardcoding a value. | Input | Type | Default | Description | |---|---|---|---| | Image | IMAGE | — | Source image — its current width/height are read directly, no separate Get Image Size node needed | | Multiplier | FLOAT | 2.0 | Multiply width and height by this factor | | Round to multiple of | INT | 8 | Snap dimensions to this multiple | **Outputs:** Width, Height, Resolution (longest side as INT → directly into SeedVR2) --- ## Example workflows Ready-to-load workflows are included in [`example_workflows/`](example_workflows): - **`smart_prompt_controller_demo.json`** — Smart Prompt Controller + Smart Profile Switch driving 4 prompt categories with auto-switching resolution and filename label - **`seedvr2_smartresize_demo.json`** — Smart Resolution Multiplier → SeedVR2 upscale → Smart Resize, a clean pattern for upscaling without hardcoding target resolution - **`smart_batch_resize_demo.json`** — minimal Smart Batch Resize setup with a Note explaining outputs and settings --- ## Why these nodes? - ComfyUI's built-in `Resize Images by Longer Edge` [BETA] has no Lanczos and no `multiple_of` snapping - `JWImageResizeByLongerSide` (comfyui-various) has no Lanczos in the official release - No existing node combines batch folder loading + longest-side resize + original filename preservation - No existing node outputs a ready-to-use `resolution` INT for SeedVR2 - No existing node cycles through multiple prompt lists with automatic list switching and an index - No existing node switches label + width + height together per category in one place - Smart Prompt Controller + Smart Profile Switch handle **uneven category sizes automatically** — 5 headshot prompts and 10 fullbody prompts cycle correctly without padding lists to match or manually tracking where one category ends and the next begins --- ## Installation **ComfyUI Manager:** search for `ComfyUI-CraftKit` and install. **Manual:** ```bash cd ComfyUI/custom_nodes git clone https://github.com/CraftopiaStudio/ComfyUI-CraftKit ``` Restart ComfyUI. All nodes appear under **CraftKit** in the node menu: - `CraftKit` → **Smart Prompt Controller 📋** - `CraftKit` → **Smart Profile Switch 🧩** - `CraftKit` → **Smart Resize 📐** - `CraftKit` → **Smart Batch Resize 📁** - `CraftKit` → **Smart Resolution Multiplier 📏** --- ## Requirements Pillow, NumPy, PyTorch — all included with ComfyUI. No extra dependencies.