One-Video Sprite Sheet Pipeline

This workflow is basically a “one-click sprite/animation factory” built around NanoBananaEdit_floyo — and it’s designed to output clean, production-ready frames (with masks + consistent sizing) instead of “pretty but messy” generations.

What this workflow does (at a glance)

It takes an input image (or batch), runs it through Nano Banana to generate a 4-frame attack animation sprite sheet, then automatically:

• splits the 2×2 sheet into individual frames

• builds masks (chroma / alpha / bounding boxes)

• crops + pads + resizes frames consistently

• composites layers with masked merges (so edges stay clean)

• re-batches frames into the right order

• and can export as a video (or a final processed batch), depending on your output path

The “big win”: reliable keying + consistent framing

Most sprite workflows fall apart in two places:

1. backgrounds aren’t truly uniform, so keying is annoying

2. frames don’t align, so the character jitters between frames

This workflow solves both by enforcing:

• hard chromakey green background (hex #00FF00) in the NanoBanana prompt

• a full chain of masking → bounding box → crop/pad → resize

• repeated ImageCompositeMasked steps to keep silhouettes clean and centered

So your output frames come out ready to drop into a game engine or timeline.

What the node chain is doing (in practical terms)

Here’s the “story” of the pipeline:

1) Generation (the creative brain)

• NanoBananaEdit_floyo generates a 2×2 grid sprite sheet with strict rules:

  • no green on the subject

  • no gradients/shadows in the background

  • no black borders between frames

  • consistent character design across all 4 frames

2) Frame extraction (turning sheet → frames)

• SplitImageWithAlpha / ImageFromBatch / RepeatImageBatch

  • slices the sheet into 4 usable frames

  • ensures each frame is treated consistently downstream

3) Mask + alignment (the “no jitter” part)

• ColorToMask + InvertMask + MaskBoundingBox+

  • isolates the character cleanly (by chroma / alpha logic)

  • finds the tight bounding box

• ImageCrop + ImagePadForOutpaintMasked + GetImageSize

  • normalizes framing so each frame is aligned

  • pads where needed so you don’t clip motion/extremities

4) Cleanup + compositing (edges and polish)

• ImageCompositeMasked (multiple stages)

  • merges layers using masks so edges don’t get crunchy

  • keeps background handling predictable

5) Output packaging

• BatchImagesNode / CreateVideo

  • reassembles frames into a batch or timeline-ready output

  • optionally exports a video preview so you can review motion fast

The workflow is now cleaner with Set/Get nodes

I updated the workflow so the mask-splitting reroutes are replaced with SetNode/GetNode pairs:

• SetNode stores each mask once (mask_1, mask_2, mask_3)

• GetNode pulls those masks anywhere they’re needed downstream

What that gives you:

• easier debugging (you can tap a GetNode anywhere and preview that exact mask)

• less spaghetti wiring

• safer edits (you won’t break three branches when you adjust one mask path)

Where this shines (real use cases)

• Game sprites: consistent frame alignment + clean silhouettes

• Storyboard panels: swap the prompt and you get a repeatable “panel → cleanup → output” system

• Content packs: batch input characters and mass-generate attack animations quickly

• Fast iteration: tune mask thresholds/crop logic once, and everything stays stable