V2V Shotboard Easy - Wan Animate 2 Edition

A native Wan-Animate-2 workflow is a big graph: loader, conditioning, context windows, chunking, pose cache, compositing, and every knob is a separate widget scattered across the canvas. IAMCCS_V2VShotboardEasyWanAnimate2 is the opposite approach - one full-screen control panel with every setting you'd actually touch, mapped to the backend graph underneath. You get the full Wan-Animate-2 motion-transfer pipeline behind a single node that hands a cine_linx plan to a companion backend.

What it controls

The panel mirrors the real Wan-Animate-2 knobs one-for-one. Source side: source_video_path, source_image_path, duration_seconds, fps, trim_start_s/trim_end_s, and frame_load_cap. Generation side: generation_width/height (832×480 default), chunk_frames (81 - the official Wan-Animate-2 default, enforced to 4n+1), generation_steps/cfg/seed with seed_mode fixed-or-increment, reference_image_strength, and the pose block (pose_strength, pose_start_percent, pose_end_percent).

Then the long-form machinery that makes this worth using over the raw nodes:

• enable_context_windows with context_length_latents (21), context_overlap_latents (8), context_schedule and context_fuse_method - the chunk-to-chunk continuity system.
• enable_pose_cache with cache_device (cpu/gpu) and cache_dtype (int8/int4/default) - so you don't re-encode pose embeddings for every chunk.
• reference_background_mode (keep_reference_background vs isolate_character, which needs reference_character_mask).
• output_background_mode - native_generated, or composite over source_video, reference_image, or a custom_background plate.
• live_chunk_preview, empty_cache_each_chunk.

Plus the text stack (global_prompt, pose_prompt, negative_prompt), model/LoRA/CLIP/VAE name slots (default "auto / graph" - the backend decides), and timeline_data for shotboard-style planning.

How it wires up

The single output is cine_linx plus a report. That contract is consumed by IAMCCS_WanAnimate2ShotboardBridge, which unpacks the plan into typed values (trim_start_s, target_frames, chunk_frames, generation_steps, pose_strength, the context settings, prompts, and an effective_lora_strength that is only nonzero when apply_distill_lora is on). Those values drive the actual IAMCCS_WanAnimate2Extends sampling node. So the Easy shell is a planner: it does the duration/trim/chunk arithmetic (chunk_count from ceil((frame_cap-1)/(chunk-1))), packages your decisions, and the backend executes them. It never loads the model itself.

The honest take

This is a quality-of-life node for people who run Wan-Animate-2 long-form a lot. If you already have a tuned native graph and know it, the Easy shell is a convenience, not a capability bump - you can reproduce every one of these settings in the raw nodes. What it genuinely buys you: one surface to edit, prompts that travel with the plan, chunk/trim arithmetic that can't go stale, and a cine_linx contract that makes the workflow printable and versionable. Where people stumble: leaving chunk_frames off a 4n+1 boundary (the panel clamps it, the raw graph doesn't), and forgetting the shell is a planner - the model must come from the backend graph you wire it into.

Install the pack via ComfyUI Manager (search "IAMCCS") or cd ComfyUI/custom_nodes && git clone https://github.com/IAMCCS/IAMCCS-nodes.git, restart.