Overview
This is a skill that calls the video generation tools of Higgsfield MCP. It selects the most suitable one of the 11 official video models for the user's intent, and handles everything from video presets and character consistency to asynchronous job polling.
Trigger keywords
Higgsfield video, Sora 2, Google Veo 3, Kling 2.1 Master, Kling 2.5 Turbo, Kling 3.0, Kling Avatars 2.0, Seedance 2.0, Seedance Pro, Cinema Studio 3.5, MiniMax Hailuo 02, Wan 2.5, UGC video, Unboxing video, Product review, AI video generation, advertising video
Prerequisite — MCP registration
moai-media When installing the plugin, .mcp.jsonautomatically registers this Higgsfield MCP (shared with the image skill). One-time OAuth authentication in Cowork just before the first call.
11 official video models — at a glance
higgsfield.ai official page as explicitly stated (2026-05).
General video (text-to-video / image-to-video)
| Model | Strengths | When suitable |
|---|---|---|
| Sora 2 ★ | Realistic, natural motion | General advertising and narrative |
| Google Veo 3 | High quality, diverse tones | Marketing and brand videos |
| Kling 2.1 Master | Person and character motion (1st-generation Master) | Person-focused |
| Kling 2.5 Turbo | Fast processing, medium quality | Quick drafts and iteration |
| Kling 3.0 ★ | Latest Kling, high quality | Emphasis on people and facial expressions |
| Seedance 2.0 | Dynamic motion | Dance, action, liveliness |
| Seedance Pro ★ | Seedance high-quality variant | Advertising and high-quality motion |
| MiniMax Hailuo 02 | Precise camera movement | Tracking, zoom, pan |
| Wan 2.5 | Diverse tones, latest from China | Experimental and new tones |
Cinematic film look
| Model | Strengths |
|---|---|
| Cinema Studio 3.5 ★ | Film grading, film look, camera directing |
Person and character consistency
| Model | Strengths |
|---|---|
| Kling Avatars 2.0 | Character consistency videos and avatar series |
6 video presets
higgsfield.ai/skills Official video presets:
| Preset | Purpose |
|---|---|
| UGC | General user-generated content tone (naturalness) |
| Unboxing | Unboxing and product opening |
| Product review | Product review (feature emphasis) |
| Hyper motion | Fast motion and dynamic |
| TV spot | Advertising TV tone (cinematic) |
| Wild Card | Experimental and creatively free tone |
If you specify a preset, an appropriate model is automatically selected and cinematic settings are applied.
Workflow
Step 1 — Identify intent, length, and purpose
| Item | Automatic inference |
|---|---|
| Length | "Short"·"5 seconds" → Kling 2.5 Turbo·Seedance / "Ad"·"8-10 seconds" → Veo 3·Sora 2 |
| Purpose | Ad · SNS · Narrative · Demo·Tutorial · Character speech |
| Input | text-to-video · image-to-video · character series |
| Preset | UGC·Unboxing·Product review·Hyper motion·TV spot·Wild Card |
If insufficient, 1 round of AskUserQuestion.
Step 2 — Automatic model·preset selection
| User expression | Automatic |
|---|---|
| "Realistic", "Natural", "General ad" | Sora 2 or Google Veo 3 |
| "High-quality brand video" | Google Veo 3 |
| "Facial expression·acting important" | Kling 3.0 or Kling 2.1 Master |
| "Quickly multiple drafts" | Kling 2.5 Turbo (Turbo processing) |
| "Dynamic motion·dance" | Seedance 2.0 or Seedance Pro |
| "Ad high quality" | Seedance Pro · Cinema Studio 3.5 · Google Veo 3 |
| "Precise camera movement" | MiniMax Hailuo 02 |
| "Cinematic look·film grading" | Cinema Studio 3.5 |
| "Character series·avatar" | Kling Avatars 2.0 |
| "Experimental·new tone" | Wan 2.5 |
| "Unboxing·review" | Product review·Unboxing preset |
| "UGC tone naturally" | UGC preset |
| "Ad TV spot" | TV spot preset + Cinema Studio 3.5 |
| "Dynamic·rapid change" | Hyper motion preset |
Step 3 — Input preparation
Text-to-video
Only prompt needed.
Image-to-video
Start image needed. The start frame determines the resulting tone.
[권장] higgsfield-image (Soul 2.0·Soul Cinema·Flux Kontext)로 시작 이미지 → image-to-video
Character series
- Use Kling Avatars 2.0
- Register character reference in advance
Step 4 — Parameter design
Common parameters
| Parameter | 값 |
|---|---|
prompt |
Action·mood·detail description |
model |
Model identifier (e.g., sora_2, veo_3, kling_3_0, seedance_pro) |
preset |
UGC · Unboxing · Product review · Hyper motion · TV spot · Wild Card |
quality |
turbo (fast·cheap) · standard · high |
aspect_ratio |
16:9 · 9:16 (Reels·short-form) · 1:1 · 4:5 |
duration_seconds |
Varies by model (5-15 seconds) |
seed |
For reproduction |
Image-to-video additional
| Parameter | 값 |
|---|---|
image_url or input_images[] |
Start image URL |
Step 5 — MCP call
ToolSearch(query: "select:mcp__higgsfield__generate_video")
mcp__higgsfield__generate_video({
model: "veo_3",
prompt: "[행동·분위기 묘사]",
aspect_ratio: "16:9",
duration_seconds: 8,
quality: "standard",
preset: "TV spot", // 옵션
image_url: "..." // image-to-video 시
})
Model identifiers are in underscore_separated format on the official page (e.g., sora_2, veo_3, kling_3_0, kling_2_5_turbo, seedance_pro, cinema_studio_3_5).
Step 6 — Asynchronous job polling
Video takes longer than images.
| Status | Average time |
|---|---|
queued |
Immediately if balance is normal |
in_progress |
Sora 2/Veo: 20-60초 · Kling Turbo: 10-25초 · Cinema Studio·Seedance Pro: 30-90초 |
completed |
Receive video URL |
failed |
Prompt·model compatibility issue |
nsfw |
Content filtering |
For long videos (15 seconds), 1-2 minutes is normal.
Step 7 — Result review·iteration
| User response | Follow-up action |
|---|---|
| "Good" | Download·SNS upload |
| "Unnatural motion" | Change model to Kling 3.0 or Veo 3 |
| "Tone is weak" | Cinema Studio 3.5 + TV spot preset |
| "Faster / shorter" | Reduce duration + Kling 2.5 Turbo or Seedance |
| "Add voice" | audio-gen (ElevenLabs) → external editing |
Usage examples
Example 1 — 8-second ad video (cinematic)
요청: "신제품 가죽 지갑 광고 영상, 시네마틱 톤, 8초"
플로우:
1. higgsfield-image (Soul Cinema)로 시작 이미지 1장
2. higgsfield-video
- model: cinema_studio_3_5 또는 veo_3
- preset: TV spot
- image_url: 1번 결과
- duration: 8s
- aspect_ratio: 16:9
- quality: high
3. 60초 대기 → 8초 영상
Example 2 — Short-form for Reels (UGC)
요청: "5초 인스타 릴스 영상, 자연스럽게"
자동: model: kling_2_5_turbo (빠름·저비용)
preset: UGC
aspect_ratio: 9:16
duration: 5s
quality: turbo
Example 3 — Character series video
요청: "브랜드 마스코트가 점프하는 5초 영상"
자동: model: kling_avatars_2_0
입력: 마스코트 이미지 reference
duration: 5s
Example 4 — Product unboxing video
요청: "신제품 언박싱 영상 10초"
자동: model: seedance_pro (제품 광고 적합)
preset: Unboxing
duration: 10s
Example 5 — Dynamic dance video
요청: "댄스 영상 짧게, 빠른 모션"
자동: model: seedance_pro
preset: Hyper motion
duration: 5s
quality: standard
Example 6 — Camera tracking video
요청: "제품을 회전하면서 보여주는 영상"
자동: model: minimax_hailuo_02 (카메라 무브먼트 정교)
duration: 6s
Output format
## Higgsfield 영상 생성 결과
### 호출 정보
- 모델: [예: veo_3]
- 프리셋: [TV spot 등 또는 없음]
- 입력: text-to-video / image-to-video
- 길이·비율·품질: [duration · aspect_ratio · quality]
- Job ID: [Higgsfield job ID]
- 소요 시간: [실제]
### 결과
- 영상 URL: [Higgsfield CDN]
- 다운로드 경로: [로컬]
- 썸네일: [첫 프레임]
### 검수
- 모션 자연스러움·의도 부합·길이·비율: [점수]
### 후속 추천
- 음성 추가 → audio-gen
- 편집·자막 → 외부 도구
- 시리즈 영상 → 같은 모델·다른 입력
Cost management
| Model | Relative cost |
|---|---|
| Kling 2.5 Turbo | Cheapest·fastest |
| Seedance 2.0 / Wan 2.5 | Cheap |
| Kling 2.1 Master / MiniMax Hailuo 02 | Medium |
| Google Veo 3 | Medium-expensive |
| Sora 2 | Expensive |
| Kling 3.0 | Expensive |
| Seedance Pro / Cinema Studio 3.5 | Expensive |
| Kling Avatars 2.0 | Expensive (special) |
Workspace balance: higgsfield.ai → Billing.
Precautions
Do
- For the first attempt, Kling 2.5 Turbosecure a quick draft with → final generation with Veo 3·Sora 2
- Image-to-video depends on the quality of the starting image for the result →
higgsfield-imageprioritize - For ads, TV spot preset + 16:9
- For Reels·short-form, 9:16 + UGC or Hyper motion preset
- For character series, use Kling Avatars 2.0
Don't
- Requesting overly complex motion in one video
- Abrupt changes in videos longer than 10 seconds
- nsfw·sensitive content
- Videos that may infringe portrait rights
- Copyrighted characters·logos
Troubleshooting
| Symptom | Cause | Solution |
|---|---|---|
| "Not connected" | OAuth unauthenticated | Cowork → Settings → MCP → Higgsfield → Connect |
queued stuck |
Insufficient balance | Top up at higgsfield.ai/billing |
failed (image-to-video) |
Starting image URL error | Generate a new image with higgsfield-image and retry |
| Unnatural result | Unsuitable model | Switch Sora 2 ↔ Veo 3 ↔ Kling 3.0 |
| Video is too short | duration not specified | Specify duration_seconds |
| Cost higher than expected | Using an expensive model | Explore first with Kling 2.5 Turbo |
| Preset effect not visible | Incompatible with model | Use the preset alone or a recommended model |
Related skills
| Skill | Timing |
|---|---|
moai-media:higgsfield-image |
Prerequisite: generate starting image·character reference |
moai-media:audio-gen |
Auxiliary: generate voice·BGM to add to the video |
moai-content:landing-page |
Follow-up: place the video on the landing page |
moai-marketing:campaign-planner |
Auxiliary: campaign-level video series |
moai-commerce:detail-page-image |
Auxiliary: product detail page video |
moai-commerce:live-commerce |
Supplementary: Live Commerce |
References
references/dop-motions.md— Video Presets and Camera Directing Guide- Higgsfield Official Site — List of 11 Models
- Higgsfield MCP Guide
- Higgsfield Skills — 6 Video Presets
Model List Source
The 11 video models and 6 video presets in this skill are higgsfield.ai official site as explicitly stated (2026-05). Models and presets may be added or changed according to Higgsfield policy.