Higgsfield Ecommerce Ad

Higgsfield E-Commerce Ad Skill

What this skill does

Crafts production-ready e-commerce and DTC advertisement prompts and routes them to the best Higgsfield model for high-converting output. On request, it assembles the full parameter set, surfaces the credit cost, and — after explicit user confirmation — generates the asset via the opt-in generation flow. Text-to-video is the default path; image-to-video is conditional on the user supplying a reference image. Product stills route to ms_image via higgsfield:generate_image.

---

Model routing

Primary and fallback video models

| | marketingstudiovideo (primary) | seedance20 (fallback) | |---|---|---| | Rationale | Dedicated marketing/ad pipeline; optimised for conversion formats | Genre/mode control + video/audio reference roles | | Resolution | 480p, 720p, 1080p | 480p, 720p, 1080p (default 720p) | | Aspect ratios | auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | auto, 21:9, 16:9, 4:3, 1:1, 3:4, 9:16 | | Duration | 4–15 s (continuous range) | 4–15 s (continuous range) | | Extra params | generate_audio bool; optional folder_id, width, height, hook_id, setting_id, ad_reference_id | mode {std, fast}; genre {auto, action, horror, comedy, noir, drama, epic} | | Media roles | avatars, image, start_image, end_image | image, start_image, end_image, video, audio |

Routing rule: Use marketing_studio_video by default — it is the dedicated ad pipeline and supports audio generation via generate_audio. Fall back to seedance_2_0 only when (1) the primary model is unavailable, or (2) the user needs genre-tagged output or a video/audio reference role not available on the primary. If falling back, inform the user of the switch. Never switch silently.

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](references/model-specs.md).

Per-platform aspect ratio

Quick rule: 9:16 for TikTok Shop / Reels / Shorts; 16:9 for YouTube and Amazon; 1:1 for Instagram feed and Pinterest. Full per-platform table: [references/model-specs.md](references/model-specs.md).

---

Still-image path (DTC Ads — ms_image)

For a product still, hero shot, or mood-board frame before committing to video — or when the user wants a high-quality static DTC product image — route to ms_image via higgsfield:generate_image.

ms_image requires a style_id — this field is REQUIRED and must be provided before generation. brand_kit_id is optional. Resolution options: 1k, 2k, 4k (the 4k enum is a legitimate option for this image model only — it does not apply to any video model in this plan). See full parameter table: [references/model-specs.md](references/model-specs.md).

After generating a still with ms_image, the output image can optionally be passed as a start_image reference to marketing_studio_video to carry the composition into video.

---

Prompt-building workflow

1. Gather intent — Confirm: product category (fashion, beauty, electronics, food,

jewelry, home, fitness…), request type (product-only, lifestyle, unboxing, DTC still), target platform, aspect ratio, desired duration, and visual tone.

1. Select model — Apply the routing table above. Choose marketing_studio_video

by default; route product-still requests to ms_image via generate_image; select seedance_2_0 only when a routing trigger applies.

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

• Opening 2-second hook from [references/hooks.md](references/hooks.md)
• Product staging, lighting, value communication, lifestyle context from

[references/ad-craft.md](references/ad-craft.md)

• Worked example prompts by product category and platform from

[references/examples.md](references/examples.md)

1. 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 is made. This skill never auto-generates.

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

This skill's primary model: marketing_studio_video

Audio: marketing_studio_video supports a generate_audio boolean. Surface this option to the user — enabling it may affect credit cost.

Media upload: Conditional — only when the user supplies a start_image, end_image, or other reference for image-to-video. This skill is NOT a mandatory-I2V skill; T2V is the default path. When a reference is provided, run the higgsfield:media_upload → higgsfield:media_confirm atomic pair first.

Tool signatures: [../../shared/mcp-tools.md](../../shared/mcp-tools.md)

---

Reference materials

| File | Contents | |---|---| | [references/model-specs.md](references/model-specs.md) | Per-model parameter tables for marketingstudiovideo, seedance20, and msimage (including its legitimate 4k resolution enum and required styleid) | | [references/ad-craft.md](references/ad-craft.md) | DTC/e-commerce ad craft: product staging, lighting, multi-angle techniques, lifestyle integration, text/CTA guide, quick-cut patterns, sound design | | [references/hooks.md](references/hooks.md) | 2-second hook framework with 12 proven e-commerce hooks and prompt phrasing templates | | [references/examples.md](references/examples.md) | Master template + 5 full example prompts by product category (fashion, skincare, electronics, food/beverage, jewelry) |

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: pixelab-ch
• Source: pixelab-ch/higgsfield-skills
• License: MIT

Install and usage instructions live in the source repository linked above.