What problem does it solve? AI video models handle audio inconsistently — lip-sync drifts, uploaded tracks get overridden by generative audio, and prompts without audio direction produce silent or mismatched sound. This Skill provides model-specific rules for writing audio prompts that produce synchronized dialogue, sound effects, ambience, and music in Higgsfield-generated video. ## Core Features & Use Cases - Four-layer audio prompting: Structures dialogue, SFX, ambient, and BGM cues with per-model syntax for Kling 3.0, Seedance 1.5 Pro/2.0, Veo 3/3.1, and Grok Imagine Video. - Lip-sync discipline: Enforces 3–8 second dialogue clips, single-face framing, locked cameras, and per-language word budgets to prevent desync. - Audio-as-conditioning: Explains Seedance 2.0 @Audio1 reference usage for beat sync, timestamped [AUDIO: Xs] script blocks, and scoped melody/voice references. - Use Case: A creator generating a 15-second Seedance 2.0 clip uploads a trimmed MP3 hook as @Audio1, anchors it with a timestamp phrase, and writes a [AUDIO: Xs] dialogue block so the character's lines lip-sync while cuts land on the beat. ## Quick Start Ask the assistant to write a Higgsfield video prompt with dialogue, ambient sound, and beat-synced music for a specific model such as Seedance 2.0 or Kling 3.0.
