modbender UGC Campaign Pipeline
Complete workflow for creating UGC-style promotional videos from a product.
Pipeline Overview
┌─────────────────────────────────────────────────────────────────────┐
│ UGC CAMPAIGN PIPELINE │
├─────────────────────────────────────────────────────────────────────┤
│ │
│ STEP 1: HERO IMAGE │
│ ├─ Input: Product image + model selection │
│ ├─ Tool: morpheus-fashion-design │
│ └─ Output: ~/clawd/outputs/{project}/morpheus/hero.png │
│ │
│ STEP 2: VARIATIONS │
│ ├─ Input: Hero image │
│ ├─ Tool: multishot-ugc │
│ └─ Output: ~/clawd/outputs/{project}/multishot/*.png (10 images) │
│ │
│ STEP 3: SELECTION │
│ ├─ Analyze all 11 images │
│ ├─ Criteria: variety, no errors, lip-sync friendly │
│ └─ Output: 4 best images selected │
│ │
│ STEP 4: SCRIPT │
│ ├─ Write 4-scene dialogue script │
│ ├─ Format: PURE DIALOGUE (no annotations) │
│ └─ Output: 4 lines of dialogue │
│ │
│ STEP 5: UGC VIDEOS │
│ ├─ Input: 4 images + 4 script lines │
│ ├─ Tool: veed-ugc (run 4 times) │
│ └─ Output: ~/clawd/outputs/{project}/ugc/*.mp4 (4 videos) │
│ │
│ STEP 6: FINAL EDIT │
│ ├─ Input: 4 videos + logo │
│ ├─ Tool: Remotion │
│ ├─ Add: subtitles, transitions, logo ending │
│ └─ Output: ~/clawd/outputs/{project}/final/video.mp4 │
│ │
└─────────────────────────────────────────────────────────────────────┘
Execution Checklist
Before Starting
- Product image received
- Brand/product name known
- Target audience understood
- Tone defined (casual, professional, energetic)
Step 1: Hero Image (Morpheus)
# Select model from catalog
cat ~/clawd/models-catalog/catalog/catalog.json | jq '[.talents[] | select(.gender == "male/female") | {id, name, ethnicity}]'
# Generate hero image
COMFY_DEPLOY_API_KEY="..." uv run ~/.clawdbot/skills/morpheus-fashion-design/scripts/generate.py \
--product "product.jpg" \
--model "models-catalog/catalog/images/model_XX.jpg" \
--brief "..." \
--target "..." \
--aspect-ratio "9:16" \
--output "~/clawd/outputs/{project}/morpheus/hero.png"
Step 2: Variations (Multishot)
COMFY_DEPLOY_API_KEY="..." uv run ~/.clawdbot/skills/multishot-ugc/scripts/generate.py \
--image "~/clawd/outputs/{project}/morpheus/hero.png" \
--output-dir "~/clawd/outputs/{project}/multishot" \
--resolution "2K" \
--aspect-ratio "9:16"
Step 3: Selection Criteria
Analyze all 11 images for:
- ✅ Face clearly visible (frontal or 3/4)
- ✅ Mouth not obscured
- ✅ No distorted hands/fingers
- ✅ Product visible
- ✅ Different from other selected images
Select 4 images that are:
- Different angles/perspectives
- Suitable for lip-sync
- Error-free
Step 4: Script Writing
Write 4 lines of pure dialogue:
ESCENA 1: [Opening hook - grab attention]
ESCENA 2: [Problem/benefit - relate to audience]
ESCENA 3: [Feature highlight - specific value]
ESCENA 4: [CTA/brand mention - close strong]
Rules:
- Pure dialogue only
- NO annotations: [excited], (pause), smiles
- NO stage directions
- NO tone indicators
- Just what the person says
Step 5: Generate Videos (VEED)
for i in 1 2 3 4; do
COMFY_DEPLOY_API_KEY="..." uv run ~/.clawdbot/skills/veed-ugc/scripts/generate.py \
--image "selected_image_$i.png" \
--script "Script line $i" \
--output "~/clawd/outputs/{project}/ugc/escena$i.mp4"
done
Step 6: Final Edit (Remotion)
Create Remotion project with:
- All 4 video clips sequenced
- Animated subtitles for each scene
- Logo animation at end
- Render to final MP4
Script Templates by Industry
Snacks/Food
Escena 1: Che, tenés que probar esto.
Escena 2: [Sabor/textura highlight]. Te pega de una.
Escena 3: Y mirá, no es solo [feature], tiene ese gustito que te deja queriendo más.
Escena 4: [Brand + Product]. Una vez que arrancás, no parás.
Tech/Gadgets
Escena 1: Mirá lo que me llegó.
Escena 2: Esto cambia todo. [Key feature].
Escena 3: Y lo mejor? [Secondary benefit].
Escena 4: [Brand]. Ya no vuelvo atrás.
Beauty/Skincare
Escena 1: Ok, necesito contarte algo.
Escena 2: Este [producto] es increíble. [Result].
Escena 3: Lo uso hace [tiempo] y mirá la diferencia.
Escena 4: [Brand]. Tu piel te lo va a agradecer.
Fashion
Escena 1: Encontré mi nueva obsesión.
Escena 2: Mirá este [prenda]. [Quality/style].
Escena 3: Combina con todo y es súper [comfort/style].
Escena 4: [Brand]. Estilo sin esfuerzo.
Output Structure
~/clawd/outputs/{project}/
├── morpheus/
│ └── hero.png # Original hero image
├── multishot/
│ ├── 1_00001_.png # 10 variations
│ ├── 2_00001_.png
│ └── ...
├── ugc/
│ ├── escena1.mp4 # Individual scene videos
│ ├── escena2.mp4
│ ├── escena3.mp4
│ └── escena4.mp4
├── assets/
│ └── logo.png # Brand logo
└── final/
└── video.mp4 # Final edited video
Quality Checklist (Before Delivery)
- Hero image: product visible, model looks natural
- Variations: 4 selected are distinct and lip-sync ready
- Script: matches brand tone, pure dialogue format
- Videos: lip-sync quality is good, no artifacts
- Final: subtitles readable, transitions smooth, logo appears
- Audio: voice quality clear, timing natural