higgsfield-seedance

Writes and repairs Seedance 2.0 video-generation prompts using cinematography structure and filter diagnostics.

Updated Jul 15, 2026
One-click install
npx skills add https://github.com/executiveusa/buffer-blaster- --skill higgsfield-seedance-executiveusa
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: higgsfield-seedance
Source: https://github.com/executiveusa/buffer-blaster-/tree/main/skills/higgsfield/skills/higgsfield-seedance
Command: npx skills add https://github.com/executiveusa/buffer-blaster- --skill higgsfield-seedance-executiveusa

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? Seedance 2.0 prompts get rejected by the content filter or produce broken renders, and users burn generation credits regenerating unchanged prompts. This Skill rewrites scene descriptions into filter-safe, cinematographically structured prompts and diagnoses why a generation failed before more credits are spent. ## Core Features & Use Cases - Structured prompt authoring: Builds prompts with the six-slot formula (camera, subject, action, setting, style, lighting) or the official block scaffold for multi-shot production work, with FOV anchors, cut ladders, and positive locks. - Failure diagnosis and recovery: Distinguishes instant filter rejections from delayed render failures, catalogs named failure modes (FPS drift, action-reversal fill, orphan limbs, filler-babble dialogue), and prescribes a prompt-side counter for each. - Preflight linting: Runs prompts through a linter that checks age markers, shot counts, enum legality, and language caps against model specs before generation. - Use Case: A user's Seedance prompt keeps getting flagged. The Skill identifies it as a filter rejection, rewrites the prompt in filmmaker voice with all six slots filled, scrubs risk tokens, and validates it with the preflight linter before the next generation attempt. ## Quick Start Ask the assistant to write a Seedance 2.0 prompt for your scene description, or paste a flagged prompt and ask why it was rejected and how to fix it.

Frequently Asked Questions about higgsfield-seedance

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I write a Seedance 2.0 prompt that passes the content filter?

Describe a complete scene, not a subject, using the six-slot formula: camera movement, subject, action, setting, style, and lighting. The filter is a language model reading full-scene intent, so prompts missing three or more slots get flagged while fully legible cinematic descriptions pass.

Why does my Seedance generation fail instantly versus after a delay?

An instant fail under 10 seconds means the content filter rejected the prompt before it reached the GPU, so rewrite the voice and remove risk tokens. A delayed fail over 30 seconds means the prompt passed but rendering failed, so simplify action density or clip length instead.

How do I fix a Seedance prompt that keeps getting flagged?

Never regenerate a flagged prompt unchanged, since the same text trips the same classifier. Rewrite in filmmaker voice, generalize body-anatomy specifics, replace sensual-register adjectives with neutral ones, then run the preflight linter to catch remaining risk tokens.

Can Seedance prompts use negative instructions like 'no jitter'?

Seedance has no negative-embedding architecture, so every token is read as a positive instruction and negation lists degrade output. Use positive constraint statements instead, such as 'Face stable. Limbs anatomically natural. Consistent lighting, no flicker.'

What are the limitations of Seedance 2.0 prompt engineering?

The engine cannot track more than three characters across cuts, breaks scene geography on reflections, treats off-screen elements as nonexistent, and has no seed parameter for reproducibility. Drafts at 480p validate the prompt but not the final take, so hero shots need frame-level review.