---
title: "AI-Powered Commerce Creator Workshop — Blueprint v2"
project: ai-filmmaking-workshop
type: workshop-plan
created: 2026-04-29
version: 2.0
duration: 1 full day (8 hours incl. lunch)
audience: 6–8 participants learning to create AI-powered commercial & UGC content
philosophy: Pre-made cases, no ideation bottleneck, competition-driven learning
standalone: Lip sync module (supplementary, not core)
---

# AI-Powered Commerce Creator — Workshop Blueprint v2

## End Goal
Each participant creates:
- 1 polished 15-sec product commercial
- 1 organic 30-sec UGC-style affiliate review
- 3 social media image posts
- 1 working ComfyUI automation workflow

All using pre-made product cases — zero time lost to ideation.

---

## Workshop Structure

```
OPENING: What's Possible (15 min) — Viral showcase + Higgsfield demo

MODULE 1: AI Text Generation (45 min)
  → Case: "Sell this smart water bottle"

MODULE 2: AI Image Creation (75 min) ← Competition included
  → Case: "Coffee brand content suite"

MODULE 3: AI Video Generation (75 min)
  → Case: "Wireless earbuds 30-sec demo"

LUNCH (60 min)

MODULE 4: AI Audio Generation (45 min)
  → Case: "Skincare vs Tech — two voices, two tones"

MODULE 5: Commercial & UGC Assembly (60 min)
  → Case: "You're an affiliate marketer. Sell this."

MODULE 6: ComfyUI Workflow Builder (60 min)
  → Case: "Automate product image creation"

BONUS (standalone): Lip Sync for Commerce (30 min)
  → When talking-head content needs lip sync

CLOSING: Show & Tell + Next Steps (15 min)
```

---

## OPENING: What's Possible (15 min)

### Objective
Show participants what people are ALREADY building — set ambition, kill skepticism.

### Agenda
| Time | Activity |
|------|----------|
| 0–3 min | Viral AI content reel: 5–6 rapid-fire examples from X/Instagram |
| 3–8 min | Higgsfield Cinema Studio demo: camera angle change, mood shift, single-photo-to-video |
| 8–12 min | "By end of today, you'll make both of these" — show the polished ad + UGC review |
| 12–15 min | Workshop rules: pre-made cases, follow-along, competition, no ideation paralysis |

### Viral Content Showcase (prepare these categories)

| Category | What to Show | Why It Went Viral |
|----------|-------------|-------------------|
| **Product "living in motion"** | Perfume bottle with liquid swirling in slo-mo, light refracting | AI motion makes static products feel premium |
| **Mood-shift product ad** | Same coffee pour — first clip: dark/moody morning, second clip: bright/sunny afternoon | Higgsfield mood-shift technique. "Wait, same shot?" |
| **"Impossible camera" reel** | Drone shot pulling back from coffee bean → through grinder → into cup → steam | Camera moves humans can't do. Novelty = shares. |
| **UGC-style "I tried this"** | Phone-selfie aesthetic, person holding product, genuine reaction | Doesn't look like an ad. Trust = conversion. |
| **Before/after glow-up** | Skincare product: "Day 1" vs "Day 30" with same AI character | Transformation narrative. Universal hook. |
| **Text-becomes-reality** | AI-generated script → storyboard → final video in 30 seconds | "Wait, AI did ALL of that?" Meta-content works. |

**Sources to pull from:**
- X/Twitter: Search "AI video commercial", "Higgsfield demo", "Runway Gen-3 ad", "Kling AI product"
- Instagram: #aivideo, #aigenerated, #aicommercial, #higgsfield
- Curate 5–6 specific posts with high engagement (10K+ likes) and download/screenshot for slides

### Higgsfield Cinema Studio — Quick Demo

| Feature | What It Does | Workshop Relevance |
|---------|-------------|-------------------|
| **50+ camera presets** | Zoom, pan, dolly, drone, handheld — pick from menu | Eliminates "prompt engineering for camera" |
| **Mood/lighting shift** | Same scene → change to golden hour, neon night, moody morning | Show 3 versions of same product shot |
| **Single-photo-to-video** | Upload one product photo → animated video with camera move | Fastest path to social content |
| **Viral presets library** | Pre-built narrative structures based on high-performing content | "Don't guess what works — use what already works" |
| **Free tier** | Available for testing | Works for workshop demo |

**Competitor awareness (mention briefly):**
- **Runway Gen-3/4** — Better for complex scenes, less camera control
- **Kling AI** — Better motion physics, fewer presets
- **Pika 2.0** — Faster iteration, viral effects, less cinematic

**The key insight:** Higgsfield is purpose-built for commerce creators. Its presets encode "what went viral" into one-click templates. The other tools are general-purpose video generators that happen to be usable for commerce.

---

## MODULE 1: AI Text Generation (45 min)

### Pre-Made Case
**"You're launching a smart water bottle that tracks hydration, glows to remind you to drink, and keeps water cold for 24 hours. Price: $49. Target: health-conscious millennials."**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–8 min | Lecture: Copywriting formulas (AIDA, PAS, hook-first) | Slides |
| 8–12 min | Demo: Write 3 formats live | ChatGPT |
| 12–30 min | Exercise: Generate 3 text variants | ChatGPT |
| 30–40 min | Peer review: Read best hook aloud, group votes | — |
| 40–45 min | Why this matters: Text feeds everything else (image prompts, video scripts, VO) | — |

### Deliverables (each participant produces)

| Format | Example | Purpose |
|--------|---------|---------|
| **Instagram caption** | "I drank 8 glasses of water every day for a week. Here's what happened 👇" | Social engagement |
| **Ad headline + body** | "The Bottle That Won't Let You Forget." + 2-line description | Paid ads, product page |
| **Video script (15 sec)** | Shot-by-shot with voiceover script | Feeds into Module 3 + 5 |

### Teaching Points
- Hook-first: you have 1.5 seconds on social
- AIDA: Attention → Interest → Desire → Action
- PAS: Problem → Agitation → Solution (best for UGC)
- The script you write here becomes the VO in Module 4 and the video in Module 3

---

## MODULE 2: AI Image Creation (75 min)

### Pre-Made Case
**"Create a complete visual suite for an indie coffee brand called 'Mornings'. Cold brew, single-origin beans, minimalist packaging. Target: design-conscious coffee lovers, 25–40."**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–10 min | Lecture: Image prompt anatomy (subject, style, lighting, composition, quality tags) | Slides |
| 10–15 min | Demo: Generate a social post + UGC photo + storyboard frame live | DALL·E 3 / Bing |
| 15–40 min | Exercise: Create 3 image types | DALL·E 3 / Bing |
| 40–55 min | **COMPETITION: Reverse Prompt Engineering** | DALL·E 3 / Bing |
| 55–65 min | Show & vote: Closest match wins | Group |
| 65–75 min | Why image quality matters: thumbnail stops the scroll, storyboard guides the video | — |

### The 3 Image Types (each participant creates)

| Type | What It Is | Prompt Guidance |
|------|-----------|-----------------|
| **Social media post** | Instagram carousel-style, product-focused, on-brand colors | "Flat lay product photography, Mornings cold brew bottle on marble surface, coffee beans scattered, warm morning light, minimalist, 1:1 square, Instagram aesthetic" |
| **UGC-style photo** | "Real person" holding product, casual, authentic | "Casual photo of person holding Mornings coffee bottle, sitting at kitchen table, morning sunlight through window, candid, unposed, phone camera quality, natural skin texture, genuine smile" |
| **Storyboard frame** | Cinematic product shot for video ad | "Cinematic wide shot, Mornings cold brew pouring into glass with ice, condensation on bottle, slow motion implied, golden hour backlight, shallow depth of field, 16:9, photorealistic, commercial quality" |

### COMPETITION: Reverse Prompt Engineering

**Format:**
1. You display a pre-generated reference image on screen (product photo, specific style)
2. Participants write a prompt they think will recreate it as closely as possible
3. All prompts submitted. You run them all through DALL·E.
4. Display results side-by-side with reference. Group votes.
5. Closest match wins.

**Why this works:**
- No one "doesn't know what to make" — the target is right there
- Forces precision: every word in the prompt matters
- Teaches prompt anatomy better than any lecture
- Competition = engagement = retention

**Reference images to prepare (3 rounds, increasing difficulty):**

| Round | Image | Difficulty | Key Test |
|-------|-------|-----------|----------|
| 1 | Simple product on white background | Easy | Basic subject + style |
| 2 | Product in lifestyle setting with specific lighting | Medium | Lighting + composition descriptors |
| 3 | UGC-style with specific "imperfections" | Hard | "Unpolished" aesthetic, camera artifacts |

---

## MODULE 3: AI Video Generation (75 min)

### Pre-Made Case
**"Create a 30-second product demo for 'PulsePods' — wireless earbuds with active noise cancellation. Key features: 36hr battery, water resistant, spatial audio. Price: $129."**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–10 min | Lecture: Three video generation approaches | Slides |
| 10–18 min | Demo: Text-to-video live | Kling / Runway |
| 18–25 min | Demo: Image-to-video + Start/End Frame technique | Kling / Runway / Higgsfield |
| 25–30 min | Demo: Higgsfield camera presets (mood shift, angle change) | Higgsfield |
| 30–60 min | Exercise: Generate 3 clips using 3 different methods | Kling / Runway / Higgsfield |
| 60–70 min | Review: Best clips, what worked, what didn't | Group |
| 70–75 min | Prep for next: Export clips for Module 5 assembly | — |

### Three Video Generation Methods

| Method | How It Works | Best For | Tool |
|--------|-------------|----------|------|
| **Text-to-Video** | Prompt only. No image input. AI generates everything. | Abstract/product concepts, quick demos | Kling, Runway |
| **Image-to-Video** | Upload a keyframe image. AI animates it with motion. | Product shots, controlled starting point | Kling, Runway, Higgsfield |
| **Start + End Frame** | Upload image A (start) AND image B (end). AI fills the transition. | Before/after, transformation, reveal | Kling, Runway |

### Video Clip Assignments

| Clip # | Method | Prompt / Input |
|--------|--------|---------------|
| 1 | **Text-to-Video** | "Cinematic close-up of wireless earbuds floating in space, pulsing sound waves visible, dark background with blue light trails. Smooth rotation. 5 seconds." |
| 2 | **Image-to-Video** | Upload: UGC photo from Module 2. Prompt: "Person puts earbuds in, expression shifts to delight as music starts, natural head nod, candid moment." |
| 3 | **Start + End Frame** | Frame A: Earbuds in closed case. Frame B: Earbuds in ears. Prompt: "Case opens, earbuds float out and into ears, smooth motion, tech aesthetic." |

### Higgsfield-Specific Demo
- Load same product image
- Apply 3 different camera presets (zoom, pan, drone)
- Apply 3 different mood presets (golden hour, studio, night)
- Show: "Same product. 9 different looks. Zero prompt engineering."

### Teaching Points
- Text-to-video = fastest, least control
- Image-to-video = most control, needs good keyframe
- Start/End frame = best for transformation/reveal narratives
- Higgsfield = best for camera control without prompt expertise
- Always generate video WITHOUT audio (add audio in Module 4)

---

## MODULE 4: AI Audio Generation (45 min)

### Pre-Made Case
**"You need two completely different voiceovers for two different products. Product A: LuxeGlow serum ($89, premium skincare). Product B: BeatBuds ($49, budget tech). Voice, tone, pace, and music must match each product's audience."**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–8 min | Lecture: Voice as brand signal — premium vs casual, energy vs calm | Slides |
| 8–15 min | Demo: Clone voice, generate VO for both products | ElevenLabs |
| 15–25 min | Exercise: Generate Product A VO (luxury skincare) | ElevenLabs |
| 25–35 min | Exercise: Generate Product B VO (budget tech) | ElevenLabs |
| 35–42 min | Exercise: Generate matching music for both | Suno |
| 42–45 min | Export: 2 VO tracks + 2 music tracks → ready for Module 5 | — |

### The Two Voices Exercise

| Attribute | Product A: LuxeGlow | Product B: BeatBuds |
|-----------|-------------------|---------------------|
| **Voice style** | Warm, smooth, aspirational | Energetic, punchy, relatable |
| **Pace** | Slow, deliberate | Fast, excited |
| **Emotion tag** | "[Warm, sophisticated, intimate]" | "[Energetic, excited, casual]" |
| **Sample script** | "Your skin deserves more than hope. It deserves science." | "36 hours of battery. Zero excuses. These just work." |
| **Music prompt** | "Luxury spa ambient, soft piano, gentle strings, 70bpm, C major" | "Upbeat tech pop, punchy drums, synth bass, 120bpm, E minor" |

### Teaching Points
- Voice IS your brand. Premium skincare ≠ budget tech.
- Emotion tags are not optional — they're the difference between "robot reading text" and "actor delivering a line"
- Music tempo and key matter: fast + minor = urgency, slow + major = trust
- Audio quality > video quality. Always export 48kHz WAV.

---

## MODULE 5: Commercial & UGC Assembly (60 min)

### Pre-Made Case
**"You're an affiliate marketer for PulsePods. You need TWO content pieces: (A) A polished 15-second product commercial for TikTok/Reels. (B) A 30-second organic UGC review for the same product — 'I just tried these and honestly...'"**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–10 min | Lecture: Polished ad vs organic UGC — why you need BOTH | Slides |
| 10–15 min | Demo: Assemble the polished ad on timeline | CapCut / Resolve |
| 15–35 min | Exercise: Build your 15-sec polished ad | CapCut / Resolve |
| 35–45 min | Exercise: Build your 30-sec organic UGC review | CapCut / Resolve |
| 45–55 min | Add affiliate CTAs: where, what, when | — |
| 55–60 min | Export both pieces | — |

### Two Content Types — Same Product, Different Vibes

| Element | Polished Ad (15 sec) | Organic UGC (30 sec) |
|---------|---------------------|---------------------|
| **Opening** | Product reveal with motion graphics | "Okay so I just got these..." (casual) |
| **Visual style** | Cinematic, color-graded, smooth transitions | Phone-selfie look, slight shake, natural light |
| **Voiceover** | Professional VO, scripted | Conversational, "thinking out loud" |
| **Music** | Brand-matched, polished | Trending audio or lo-fi |
| **Text overlay** | Key features, price | "Link in bio", "Honest review ↓" |
| **CTA** | "Available now. Link in bio." | "Not sponsored. I paid for these. Link below if you want to try." |
| **Trust signal** | Brand credibility | Personal testimony |

### Affiliate CTA Formats (teach these)

| Platform | CTA Example | Placement |
|----------|------------|-----------|
| TikTok/Reels | "Link in bio 👆" | Last 2 seconds |
| Instagram Story | "Tap the link to shop" | Swipe-up sticker |
| YouTube Shorts | "Check the description" | End screen |
| All platforms | "Comment 'LINK' and I'll DM you" | Engagement bait |

### Assembly Checklist (per piece)
- [ ] Video clips in sequence
- [ ] VO aligned with visuals
- [ ] Music layered, ducked during VO
- [ ] Text overlays legible, timed correctly
- [ ] CTA clear and placed at peak attention
- [ ] Color grade: polished ad = consistent, UGC = "natural"
- [ ] Export: 1080×1920 (9:16 vertical) for TikTok/Reels

---

## MODULE 6: ComfyUI Workflow Builder (60 min)

### Pre-Made Case
**"You're a content creator who needs to produce 50 product images per week — different angles, different moods, different backgrounds. Build a workflow that does this in one click."**

### Agenda
| Time | Activity | Tool |
|------|----------|------|
| 0–10 min | Lecture: What is node-based AI? Why automation matters for commerce | Slides |
| 10–20 min | Demo: Build a product image workflow live | ComfyUI |
| 20–40 min | Exercise: Load pre-built workflow → modify → run | ComfyUI |
| 40–50 min | Show: Video workflow (Image → AnimateDiff → Save) | ComfyUI |
| 50–60 min | Discussion: Where does this lead? (API automation, batch content, template systems) | — |

### Pre-Built Workflows (prepare these as .json files)

| Workflow | What It Does | Nodes |
|----------|-------------|-------|
| **product-image.json** | Text prompt → product image on specified background | Load Checkpoint → CLIP Encode → KSampler → VAE Decode → Save |
| **product-variations.json** | Same product, 4 different backgrounds (batch) | Same as above + batch processing with seed variation |
| **product-video.json** | Static product image → animated video with camera move | Load Image → AnimateDiff → VAE Decode → Save Video |

### Teaching Points
- The manual work you did today (generate image → download → upload to video tool → download) can be ONE ComfyUI click
- Node thinking = assembly line thinking. Each box does one job.
- Start simple. Add complexity only when you need it.
- The business case: 50 images/week manually = 5 hours. With ComfyUI = 10 minutes.
- Figma Weave (cloud, polished) vs ComfyUI (local, unlimited) — same concept, different audience

---

## BONUS: Lip Sync for Commerce (30 min, standalone)

### When Commerce Content Needs Lip Sync
```
Talking-head UGC review        → HeyGen or Sync.so
AI avatar product explainer    → HeyGen
Character/spokesperson ad      → Sync.so or Runway Act-One
Product demo with host         → Sync.so
Text/image-only content        → Skip (no lips visible)
```

### Quick Demo
1. Take the UGC video from Module 3
2. Generate a VO in Module 4 voice
3. Upload both to Sync.so → download synced result
4. Compare: unsynced (voiceover) vs synced (lip-matched)

### The Rule
```
Only sync when lips are VISIBLE and the shot is CLOSE.
For everything else: voiceover works.
```

---

## Competition: Reverse Prompt Engineering (Full Rules)

### Setup
- 3 rounds, increasing difficulty
- Participants submit prompts via shared doc or chat
- You run all prompts through DALL·E 3 / Bing Image Creator
- Display results on projector, numbered, anonymous
- Group votes on closest match + most creative interpretation

### Round Structure

| Round | Time | Reference Image | What It Tests |
|-------|------|----------------|---------------|
| 1 | 10 min | Simple product on white | Subject identification, basic descriptors |
| 2 | 10 min | Product in scene with lighting | Lighting vocabulary, composition |
| 3 | 10 min | UGC-style with imperfections | "Unpolished" aesthetic, camera artifacts |

### Scoring
- Closest visual match: 3 points
- Most creative interpretation: 1 point
- Winner: Most total points across 3 rounds

### Prize
- Something small but meaningful: "Prompt Engineer Champion" badge, or first pick of next case study, or a coffee gift card

---

## Workshop Outputs (Per Participant)

| Deliverable | Module | Format |
|------------|--------|--------|
| 3 text variants (caption, ad, script) | 1 | Text file |
| 3 images (social post, UGC, storyboard) | 2 | PNG/JPEG |
| 3 video clips (text, image, start/end) | 3 | MP4 |
| 2 voiceovers + 2 music tracks | 4 | WAV |
| 1 polished ad (15 sec) | 5 | MP4, 9:16 |
| 1 organic UGC review (30 sec) | 5 | MP4, 9:16 |
| 1 ComfyUI workflow | 6 | JSON |

---

## Equipment & Materials

### Per Person
- [ ] Laptop (Mac or Windows)
- [ ] Headphones/earbuds
- [ ] Accounts created (send checklist 48hrs before)

### Room Setup
- [ ] Projector for demos + competition results
- [ ] GPU demo station (ComfyUI)
- [ ] Stable WiFi

### Organizer Materials
- [ ] Slide deck (v2 outline)
- [ ] Pre-made reference images for competition (3 rounds)
- [ ] Pre-built ComfyUI workflows (3 .json files)
- [ ] Curated viral content screenshots (5–6 posts from X/Instagram)
- [ ] Higgsfield account + pre-generated demo clips
- [ ] Cheat sheet (1 page per participant)
- [ ] Pre-workshop setup checklist (emailed 48hrs before)

---

## Pre-Made Cases Summary

| Module | Product/Case | What They Make |
|--------|-------------|----------------|
| 1 | Smart water bottle ($49) | Caption, ad copy, video script |
| 2 | Mornings Coffee brand | Social post, UGC photo, storyboard |
| 3 | PulsePods earbuds ($129) | 3 video clips (3 methods) |
| 4 | LuxeGlow serum + BeatBuds | 2 VO styles + music |
| 5 | PulsePods affiliate | Polished ad + UGC review |
| 6 | Product image automation | ComfyUI workflow |

---

*Blueprint v2. Commerce-first. Pre-made cases. Competition-driven. No ideation bottleneck.*