Higgsfield Anime Action

Higgsfield Anime-Action Skill

What this skill does

Generates anime-style video from a user-supplied anime frame or illustration using Higgsfield. The skill applies the anime visual language — cel shading, limited animation, impact frames, speed lines, color psychology — to bring the source image into motion.

A source anime frame is always required. This skill is a mandatory image-to-video (I2V) skill — it animates a user-supplied image and cannot run without one. Before any generation call can happen, the user must provide the anime frame and the higgsfield:media_upload → higgsfield:media_confirm flow must complete successfully.

Model routing

Primary and fallback video models

Routing rule: Use wan2_7 by default. Fall back to wan2_6 when wan2_7 is unavailable. If falling back, inform the user that: (1) duration becomes discrete [5, 10, 15] only — no 2, 3, 4, 6, 7, 8, 9 etc., (2) 4:3 and 3:4 aspect ratios are not available, and (3) the image role changes from start_image to image. Never switch models silently.

MODEL-06 directive: If a parameter is rejected at generation time, call higgsfield:models_explore with the target model name. Full parameter tables: references/model-specs.md.

Per-platform aspect ratio and duration

Quick rule: 9:16 for TikTok / Reels / Shorts, 16:9 for YouTube, 1:1 for Instagram feed, 3:4 for portrait/manga format (wan2_7 only). Full per-platform table: references/model-specs.md.

Prompt-building workflow

1. Gather intent — Confirm with the user: anime genre (shonen, seinen, magical girl, mecha, slice-of-life, etc.), target platform, aspect ratio, duration, and any specific visual effects or effects style.

2. Select model — Apply the routing table above. wan2_7 by default.

3. Build the prompt — Use the craft principles in the references below:

   • 2-second hook technique from references/hooks.md
   • Anime visual language, action choreography, style transfer, camera techniques from references/anime-craft.md
   • Worked example prompts by anime genre from references/examples.md

4. Present for review — Show the assembled prompt and all parameters to the user for review and refinement before any upload or generation call.

Opt-in generation

Generation costs Higgsfield credits and requires explicit user confirmation. This skill never auto-generates.

Full step-by-step flow (confirmation gate, balance/cost surface, generate → poll → display): ../../shared/generation-flow.md

This skill's primary model: wan2_7

Media upload — MANDATORY (not conditional)

This skill always requires a source anime frame. The media branch is not conditional — it always runs before generation. Do not skip it.

Required sequence:

1. Ask the user to provide the anime frame or illustration file (PNG, JPG, or WebP).
2. Call higgsfield:media_upload with the file → receives a pending_id.
3. Call higgsfield:media_confirm with the pending_id → receives a confirmed_id.
4. Attach the confirmed_id to the input_files array before calling higgsfield:generate_video:
   • wan2_7 (primary): role = start_image
   • wan2_6 (fallback): role = image

// wan2_7 (primary)
input_files: [{ "id": "<confirmed_id>", "role": "start_image" }]

// wan2_6 (fallback)
input_files: [{ "id": "<confirmed_id>", "role": "image" }]

Never pass a pending_id directly to input_files. Never attempt generation without first completing the upload pair. wan2_6's image role is REQUIRED — a missing or unconfirmed image will cause the request to be rejected.

See ../../shared/generation-flow.md Step 2b for the atomic-pair detail and full input_files structure.

Tool signatures: ../../shared/mcp-tools.md

Reference materials