Organic Fill Animation 🌿

Eden_OrganicFillAnimation turns an image into a video of a shape growing into being. It converts your image to a mask, drops one or more "seeds" on it, and animates the mask spreading outward from those seeds with organic, noise-driven growth - the effect looks like a liquid, a plant, or a stain slowly filling the shape. It's a procedural animation generator: no video model, no sampler, just a growth simulation producing frames.

How it works

The core is a cellular-growth simulation. Your image gets reduced to a threshold mask - pick color_channel (luminance, red, green, blue, or random) and a threshold_value (default 128), and pixels below the threshold count as "fillable area." A seed of seed_radius pixels is placed (position chosen by starting_position - center, corners, edges, or random - with position_randomness jittering it). Each step, the boundary pixels are tested against a noise field: a pixel grows only where the noise value exceeds growth_threshold. noise_low and noise_high define that noise field's range and are the main "how wild is the growth" knobs. island_connection_radius lets detached regions get connected back to the main blob; active_region_padding bounds how far the growth can wander.

The output is a frame sequence resampled to hit target_seconds at your fps, with a hold on the final frame (hold_final_frame_fraction) and optional loop (which plays the growth in reverse back to the start for a seamless cycle). num_seeds (1–10) grows from several points at once. invert_input flips which pixels are fillable.

Inputs you'll actually touch

• image (IMAGE) - the input.
• threshold_value, color_channel - what counts as fillable.
• growth_threshold + noise_low/noise_high - the organic feel.
• starting_position, num_seeds - where growth starts.
• target_seconds, fps, loop - the video's shape.

Output: frames (IMAGE, a batch of frames ready for a video encode).

Where it fits

It's a VFX gadget for animated transitions and generative-video scaffolding. Feed it a logo mask and you get an organic logo-reveal. Feed it a depth-based mask and you get a fill animation that pairs with image-to-video. It slots in front of a video encoder, or the frames can be fed back into ComfyUI as a conditioning reference. It's niche, and honestly that's fine - it does one distinctive thing well and it's GPU-cheap (CPU simulation, mostly torch tensor ops).

Installing it

Part of the Eden.art nodesuite. Install via ComfyUI Manager (search "Eden.art nodesuite") or git clone https://github.com/edenartlab/eden_comfy_pipelines into custom_nodes/, then pip install -r requirements.txt and restart.

Common issues

Growth can stall if growth_threshold is too high - the simulation just stops filling. If a run finishes instantly with a mostly-empty result, drop growth_threshold. The noise range defaults (0.0–1.2) are a safe starting band. And remember: this generates frames, not a video file - you still need a video-save/encode node downstream, and long durations at high fps produce a lot of frames to hold in memory.