Batch Processing

Batch processing lets you run an entire folder of images (or a manually assembled collection) through your graph automatically. Load, filter, color-correct, resize, and export hundreds of images without touching each one individually.

How It Works

1. Create a batch source: Image Folder loads from a directory on disk. Collect bundles hand-picked images. Palette Remap generates variants from a palette.
2. Wire the batch output into any downstream node (Blur, Levels, Resize, Sharpen, or anything else).
3. The graph evaluates every frame through the pipeline automatically. No special batch mode needed.
4. Connect Export Batch at the end to write every processed frame as a named file.

Transparent Frame Propagation

The key concept: any node that accepts an image input also accepts a batch. When a batch reaches a non-batch node, the graph runs that node once per frame and re-bundles the results automatically. You don’t need batch-specific versions of filters. Blur processes a batch of 50 images the same way it processes one.

Input pins show the frame count when fed by a batch source (e.g., image (x47)).

Batch Sources

Three node types produce batch outputs:

• Image Folder scans a directory with glob-style filtering (*.png;*.jpg), optional recursive scanning, sorting by name/date/size, and a frame limit for testing.
• Collect bundles hand-picked images from anywhere in your graph. Drag to reorder frames, and add names and color coding for organization.
• Palette Remap generates one image per palette slot. Its Batch output flows into the same downstream pipeline.

Splitting and Routing

Split decomposes a batch back into individual images or sub-batches using four modes:

• Individual: one output per frame (up to 64 outputs).
• Chunks: divide into N equal sub-batches (2 to 16).
• Slice: extract a contiguous range by start index and count.
• Filter: route frames by glob pattern on their names (e.g., hero_*).

Output pins rebuild dynamically as you switch modes or the incoming batch changes.

Export and Naming

Export Batch writes every frame as a separate file. The filename template supports tokens:

Token	Expands to
{name}	Frame name (palette slot name or Collect label)
{source}	Source filename (from Image Folder)
{index}	Frame number (zero-padded)
{width} / {height}	Output dimensions
{date}	Current date
{node}	Node label

A live preview below the template field shows the resolved filename as you edit. Export Batch supports all six output formats (PNG, JPEG, WebP, TIFF, TGA, BMP) and can pack frames into a sprite sheet with Grid or MaxRects packing, padding, trim, and edge extrusion.

AI and Batches

When a batch reaches an AI node, a confirmation modal shows the total credit cost with cache hits pre-deducted before any generation starts. Each frame processes individually. Failed frames can be retried with one click without re-running the entire batch.

Performance

Settings > Performance exposes two batch-specific controls:

• Batch concurrency cap: how many frames evaluate in parallel.
• Memory ceiling: the threshold at which frames spill to disk instead of staying in memory.

Every node in the pipeline shows a live progress bar during batch evaluation, so you can see exactly how far the run has progressed at a glance.

Common Workflows

• Asset normalization: load a folder of mixed-size icons, resize to a consistent dimension, apply sharpening, export with standardized naming.
• Batch color grading: run product photos through Levels, Color Balance, and Vibrance, then export all at once.
• Game asset pipeline: generate palette variants with Palette Remap, apply filters to each frame, export as individually named files or a sprite sheet.
• Batch background removal: wire Image Folder through AI Remove Background and Export Batch to process an entire folder of product photos.