Higgsfield Music Video

Higgsfield Music-Video Skill

What this skill does

Crafts production-ready music video prompts and routes them to the best Higgsfield model for beat-synced, audio-driven output. Handles performance videos, narrative videos, abstract visualizers, lyric videos, and multi-segment assembly for full-length songs.

An audio track is always required. This is a mandatory-audio skill: every generation needs a user-supplied audio file uploaded and confirmed via the media branch before calling higgsfield:generate_video. The confirmed audio is attached with role audio.

Model routing

Primary and fallback video models

Routing rule: Use veo3_1 by default. Fall back to veo3_1_lite only when the primary is unavailable. When falling back, inform the user of what they give up: native audio quality and the veo-3-1-preview quality tier. Never switch silently.

Aspect and duration constraints:

• Neither model supports 1:1 aspect ratio.
• Durations are the discrete set [4, 6, 8] — do not pass 5, 7, 10, or 15.
• A full music video requires multiple segments (see model-specs.md for planning guidance).

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

Prompt-building workflow

1. Gather intent — Confirm: song genre, visual strategy (performance / narrative / abstract), target platform, aspect ratio, mood/aesthetic, desired clip duration.

2. Select model — Apply the routing table above. Use veo3_1 by default.

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

   • Beat-sync techniques, energy mapping, hook framework, camera techniques: references/beat-sync.md
   • Genre-specific visual language, colors, lighting, trigger words: references/genres.md
   • Master template and worked example prompts: references/examples.md

4. Plan segments — Because veo3_1 clips are 4–8 s, map the song's structure to segments before generating. See references/model-specs.md for multi-segment planning.

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

Opt-in generation

Generation costs Higgsfield credits and requires explicit user confirmation before any generate call. 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: veo3_1

Media upload — MANDATORY (GEN-04):

An audio track is always required for video generation. Before calling higgsfield:generate_video:

1. Ask the user to provide their audio file.
2. Upload it: run higgsfield:media_upload → returns a pending_id.
3. Confirm it: run higgsfield:media_confirm → returns a confirmed_id.
4. Attach it in input_files with role audio:

   [{ "id": "<confirmed_id>", "role": "audio" }]
   

If the user also provides a reference image for the video, add it with role start_image in the same input_files array:

[
  { "id": "<audio_confirmed_id>",  "role": "audio" },
  { "id": "<image_confirmed_id>",  "role": "start_image" }
]

Never pass a pending_id directly to input_files — it will be rejected. See ../../shared/generation-flow.md Step 2b for the full atomic-pair detail.

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

Reference materials