V2V Shotboard Easy - SCAIL Edition

SCAIL-2 is the interesting one in the motion-transfer space: Z.ai's Wan-2.1 fine-tune dropped the stick-figure pose representation entirely, which is what unlocks animals, non-human characters, and multi-person identity. IAMCCS_V2VShotboardFreeScail is the IAMCCS "Easy" control shell for SCAIL-2 graphs - the same full-screen-planner philosophy as the Wan-Animate-2 edition, but tuned to SCAIL-2's defaults (16fps, 5.0625s source default, portrait 512×896 canvas) and to the thing SCAIL-2 actually cares about: single vs multi-person identity.

What it controls

Source and trim side is familiar: source_video_path, source_image_path, duration_seconds, fps (16 default - SCAIL-2's native rate), trim_start_s/trim_end_s, frame_load_cap. Generation side: generation_width/height (512×896), chunk_frames (81, clamped to 4n+1) with overlap_frames (5, also 4n+1'd), generation_steps/cfg/seed, and a LoRA slot (lora_name defaulting to "auto / graph", lora_strength).

The SCAIL-specific pair:

• scail_identity_mode - single_person or multi_person_identity. This is the big SCAIL-2 knob: without a pose skeleton, the model can hold multiple identities from reference images, and this selects the backend profile (scail2_single_person / scail2_multi_person_identity).
• show_reference_background - true keeps the reference scene visible as context; unchecking it flips replacement_mode on, i.e. the reference's background is suppressed so the character replaces rather than blends.
• interpolate_to_32fps (default on) - SCAIL-2 generates at 16fps; this routes the output through interpolation to 32fps, so the delivered timeline (scail_output_stage: final_32fps_upscaled) is the smooth master rather than the raw 16fps frames.

Plus the usual text stack (global_prompt, negative_prompt), output_prefix, timeline_data, and a preview_stage selector (source/reference/pose/mask/intermediate/output).

How it wires up

Like its Wan-Animate sibling, it's a planner: it does the trim and chunk arithmetic (segment_count = ceil(cap / (chunk - overlap))), packages everything into backend_settings plus a timeline manifest, and emits cine_linx + report. It doesn't load SCAIL-2 - "control shell for existing backend nodes in the same workflow," per the node description. You're meant to drop it in front of an existing SCAIL-2 graph and drive that graph's settings from the plan.

The honest take

The 16fps default and the 5.0625s source default are deliberate and you should keep them: SCAIL-2's own workflow runs at 16fps, and interpolating to 32fps is the intended finishing path. People who fight the defaults usually end up with juddery motion and blame the model. The multi-person mode is the genuinely useful differentiator here - on Wan-Animate-2 you'd be fighting the pose representation to do what SCAIL-2 does natively.

Ships with IAMCCS-nodes: ComfyUI Manager → search "IAMCCS", or cd ComfyUI/custom_nodes && git clone https://github.com/IAMCCS/IAMCCS-nodes.git, then restart. SCAIL-2 itself is a model (and its ComfyUI integration) outside this pack.