Generate AI video prompts: $ARGUMENTS
$ARGUMENTS should include:
- Description of the video concept (transition, reveal, transformation, loop)
- Optionally: start frame image path (to use as reference)
- Optionally: end frame image path
- Optionally: target video model (runway, kling, pika, higgsfield)
- Optionally: duration, aspect ratio
- Empty — ask the user for the concept
Before Starting
- Read
src/lib/config/tenant.config.ts for brand context
- If start/end frame images exist, read them for visual reference
- Determine the video type from the categories below
Video Types
1. Product Reveal / Deconstruction
Object assembles or disassembles — scroll-stop format.
- Start: assembled product on clean background
- End: exploded view with all components floating
- Or reverse: parts → assembled (equally compelling)
2. Transformation / Before-After
Scene or subject transforms between two states.
- Start: one state (e.g., empty room, closed book, bare landscape)
- End: transformed state (e.g., community gathering, open book with light, flourishing garden)
3. Parallax / Depth Reveal
Camera moves through layers of a scene, revealing depth.
- Static 2.5D image brought to life with subtle camera motion
- Great for hero sections, course intros
4. Zoom / Scale Transition
Macro to micro or micro to macro journey.
- Start: wide establishing shot
- End: extreme close-up detail (or reverse)
5. Loop / Cinemagraph
Seamless looping motion — a single element moves while the rest is still.
- Flickering candle, turning pages, flowing water, breathing animation
- Perfect for website hero backgrounds
6. Kinetic Typography
Text animates into frame with cinematic motion.
- Words reveal, assemble, or transform
- Great for course trailers, social clips
7. Narrative Sequence
Multi-scene storyboard — 2-4 connected shots telling a micro-story.
- Each scene described with transitions between them
Prompt Structure
Every video prompt has 5 sections. Write each explicitly for the target video model.
SECTION 1 — Start Frame
START FRAME:
[Detailed description of the opening image — subject, composition, lighting, colors, mood.
Include camera position, lens, and DOF if photorealistic.]
SECTION 2 — End Frame
END FRAME:
[Detailed description of the final image — what has changed, what remains constant.
Explicitly state what is DIFFERENT and what is the SAME as the start frame.]
SECTION 3 — Transition Choreography
TRANSITION:
[Describe the motion step by step. Include:
- What moves first, second, third (sequencing)
- Speed and easing (slow-in, slow-out, constant, accelerating)
- Direction of motion (up, down, outward, inward, along axis)
- Any rotations or reveals
- Timing: when each motion starts and ends relative to total duration
- What does NOT move (critical for stability)]
SECTION 4 — Style & Constraints
STYLE:
- Photorealistic / Cinematic / Illustrated / Abstract
- Lighting: consistent throughout, no flickering
- Camera: locked tripod / slow dolly / orbit / handheld
- Background: pure white / environmental / gradient
- Color: [palette description]
- Quality: high fidelity, smooth motion, no artifacts
SECTION 5 — Technical Specs
SPECS:
- Duration: [4-10 seconds]
- Aspect ratio: [16:9 / 9:16 / 1:1]
- Frame rate: [24fps / 30fps / 60fps]
- Resolution: [1080p / 4K]
Brand-Aligned Video Concepts for Alan Hirsch Platform
Course Trailer Concepts
- Book opening: A closed book slowly opens, pages fan, text illuminates — camera pushes in
- Community forming: Empty chairs in a circle → people arriving, settling in, conversation beginning
- Seed to tree: Time-lapse of a seed breaking soil, growing into a spreading tree (formation metaphor)
- Map unfolding: A folded map opens and spreads across a wooden table, routes illuminating
Social Content Concepts
- Quote reveal: Warm paper texture, words write themselves in ink — serif typography
- Scroll-stop deconstruction: Book or journal explodes into floating pages, binding, cover
- Candle cinemagraph: Warm scene with a single flickering candle — loopable hero
Hero Section Concepts
- Parallax neighborhood: Layers of a community street scene with subtle depth motion
- Breathing texture: Warm linen or paper texture with barely perceptible motion — alive but calm
- Golden hour shift: Subtle lighting shift from morning to golden hour across a static scene
Model-Specific Tips
Runway Gen-3/Gen-4
- Upload start and end frames as "First Frame" and "Last Frame"
- Keep prompt under 300 words for best results
- Specify camera motion explicitly: "locked tripod, no camera movement"
- Duration: 4-10 seconds
- Use "cinematic, smooth, high quality" as quality markers
Kling
- Supports longer durations (up to 15 seconds)
- Strong at character consistency
- Prompt in clear, sequential sentences
- Add "no sudden movements, smooth transitions" for stability
Pika
- Best for stylized/illustrated looks
- Shorter durations (3-5 seconds) produce better quality
- Good at text animation
- Use "consistent lighting, no flickering" explicitly
Higgsfield
- Excels at product/object animations
- Upload reference images for best consistency
- Strong at physics-based motion (floating, falling, assembling)
Execution
Step 1 — Define the Concept
Work with the user to nail down:
- What type of video (from categories above)?
- What's the subject?
- What's the start state and end state?
- What's the intended use (social, hero, course trailer)?
Step 2 — Generate Start/End Frame Prompts
If the user doesn't have start/end images yet, generate NB2 image prompts first:
- Use
/asset-generate or /asset-product-shot for the start frame
- Use
/asset-generate or /asset-exploded-view for the end frame
- Ensure both use identical lighting, camera angle, and palette
Step 3 — Write the Video Prompt
Compose the full 5-section video prompt. Show to user for approval.
Step 4 — Generate Variants
Offer 2-3 variants:
- Standard: As described (e.g., 5 seconds, forward)
- Reverse: End → Start (assembly instead of deconstruction)
- Loop: Forward → Reverse seamlessly (for website backgrounds)
Step 5 — Deliver
Present all prompts in chat. If the user wants a polished delivery page, suggest /asset-deliver.
Output Format
## Video Prompt Set: [CONCEPT NAME]
### Concept
[1-2 sentence description of the video]
### Use Case
[Where this will be used — social, hero section, course trailer, etc.]
---
### START FRAME PROMPT
[For image generation — paste into NB2/Midjourney/etc.]
{start frame prompt}
---
### END FRAME PROMPT
[For image generation — paste into NB2/Midjourney/etc.]
{end frame prompt}
---
### VIDEO TRANSITION PROMPT
[Paste into your video model, upload start frame as first frame, end frame as last frame]
{full 5-section video prompt}
---
### VARIANT — Reverse Assembly
{reverse version}
### VARIANT — Seamless Loop
{loop version}
---
### Recommended Settings
| Setting | Value |
|---------|-------|
| Duration | 5 seconds |
| Aspect | 16:9 |
| Frame rate | 24fps |
| Resolution | 1080p minimum |
| Model | Any (Runway, Kling, Pika, Higgsfield) |
### Workflow
1. Generate start frame image with `/asset-generate` using the start frame prompt
2. Generate end frame image with `/asset-generate` using the end frame prompt
3. Upload both to your video model with the transition prompt
4. Use `/asset-brand-check` on the frames before generating video
Error Recovery
| Issue |
Fix |
| Video model produces jittery motion |
Add "smooth, eased motion, no sudden jumps" and reduce complexity |
| Start/end frames look inconsistent |
Regenerate end frame using start frame as reference in NB2 |
| Motion too fast |
Increase duration, add "slow, deliberate, contemplative pace" |
| Background changes during transition |
Add "background remains static and unchanged throughout" |
| Artifacts on moving elements |
Simplify the transition — fewer moving parts, one clear axis of motion |
| Camera drifts |
Add "locked-off tripod shot, absolutely no camera movement" |
1---2name: asset-video-prompt3description: Generate video transition prompts for AI video models (Runway, Kling, Pika, Higgsfield). Creates start/end frame descriptions, transition choreography, and model-agnostic instructions. Use for scroll-stopping social content, product reveals, course trailers, or any motion asset.4user-invocable: true5allowed-tools: Read, Write, Edit, Bash, Grep, Glob6metadata:7 source: https://skillmd.com/skills/joshuashepherd/asset-video-prompt8---910Generate AI video prompts: $ARGUMENTS1112$ARGUMENTS should include:13- Description of the video concept (transition, reveal, transformation, loop)14- Optionally: start frame image path (to use as reference)15- Optionally: end frame image path16- Optionally: target video model (runway, kling, pika, higgsfield)17- Optionally: duration, aspect ratio18- Empty — ask the user for the concept1920## Before Starting21221. Read `src/lib/config/tenant.config.ts` for brand context232. If start/end frame images exist, read them for visual reference243. Determine the video type from the categories below2526## Video Types2728### 1. Product Reveal / Deconstruction29Object assembles or disassembles — scroll-stop format.30- Start: assembled product on clean background31- End: exploded view with all components floating32- Or reverse: parts → assembled (equally compelling)3334### 2. Transformation / Before-After35Scene or subject transforms between two states.36- Start: one state (e.g., empty room, closed book, bare landscape)37- End: transformed state (e.g., community gathering, open book with light, flourishing garden)3839### 3. Parallax / Depth Reveal40Camera moves through layers of a scene, revealing depth.41- Static 2.5D image brought to life with subtle camera motion42- Great for hero sections, course intros4344### 4. Zoom / Scale Transition45Macro to micro or micro to macro journey.46- Start: wide establishing shot47- End: extreme close-up detail (or reverse)4849### 5. Loop / Cinemagraph50Seamless looping motion — a single element moves while the rest is still.51- Flickering candle, turning pages, flowing water, breathing animation52- Perfect for website hero backgrounds5354### 6. Kinetic Typography55Text animates into frame with cinematic motion.56- Words reveal, assemble, or transform57- Great for course trailers, social clips5859### 7. Narrative Sequence60Multi-scene storyboard — 2-4 connected shots telling a micro-story.61- Each scene described with transitions between them6263## Prompt Structure6465Every video prompt has 5 sections. Write each explicitly for the target video model.6667### SECTION 1 — Start Frame68```69START FRAME:70[Detailed description of the opening image — subject, composition, lighting, colors, mood.71Include camera position, lens, and DOF if photorealistic.]72```7374### SECTION 2 — End Frame75```76END FRAME:77[Detailed description of the final image — what has changed, what remains constant.78Explicitly state what is DIFFERENT and what is the SAME as the start frame.]79```8081### SECTION 3 — Transition Choreography82```83TRANSITION:84[Describe the motion step by step. Include:85- What moves first, second, third (sequencing)86- Speed and easing (slow-in, slow-out, constant, accelerating)87- Direction of motion (up, down, outward, inward, along axis)88- Any rotations or reveals89- Timing: when each motion starts and ends relative to total duration90- What does NOT move (critical for stability)]91```9293### SECTION 4 — Style & Constraints94```95STYLE:96- Photorealistic / Cinematic / Illustrated / Abstract97- Lighting: consistent throughout, no flickering98- Camera: locked tripod / slow dolly / orbit / handheld99- Background: pure white / environmental / gradient100- Color: [palette description]101- Quality: high fidelity, smooth motion, no artifacts102```103104### SECTION 5 — Technical Specs105```106SPECS:107- Duration: [4-10 seconds]108- Aspect ratio: [16:9 / 9:16 / 1:1]109- Frame rate: [24fps / 30fps / 60fps]110- Resolution: [1080p / 4K]111```112113## Brand-Aligned Video Concepts for Alan Hirsch Platform114115### Course Trailer Concepts116- **Book opening**: A closed book slowly opens, pages fan, text illuminates — camera pushes in117- **Community forming**: Empty chairs in a circle → people arriving, settling in, conversation beginning118- **Seed to tree**: Time-lapse of a seed breaking soil, growing into a spreading tree (formation metaphor)119- **Map unfolding**: A folded map opens and spreads across a wooden table, routes illuminating120121### Social Content Concepts122- **Quote reveal**: Warm paper texture, words write themselves in ink — serif typography123- **Scroll-stop deconstruction**: Book or journal explodes into floating pages, binding, cover124- **Candle cinemagraph**: Warm scene with a single flickering candle — loopable hero125126### Hero Section Concepts127- **Parallax neighborhood**: Layers of a community street scene with subtle depth motion128- **Breathing texture**: Warm linen or paper texture with barely perceptible motion — alive but calm129- **Golden hour shift**: Subtle lighting shift from morning to golden hour across a static scene130131## Model-Specific Tips132133### Runway Gen-3/Gen-4134- Upload start and end frames as "First Frame" and "Last Frame"135- Keep prompt under 300 words for best results136- Specify camera motion explicitly: "locked tripod, no camera movement"137- Duration: 4-10 seconds138- Use "cinematic, smooth, high quality" as quality markers139140### Kling141- Supports longer durations (up to 15 seconds)142- Strong at character consistency143- Prompt in clear, sequential sentences144- Add "no sudden movements, smooth transitions" for stability145146### Pika147- Best for stylized/illustrated looks148- Shorter durations (3-5 seconds) produce better quality149- Good at text animation150- Use "consistent lighting, no flickering" explicitly151152### Higgsfield153- Excels at product/object animations154- Upload reference images for best consistency155- Strong at physics-based motion (floating, falling, assembling)156157## Execution158159### Step 1 — Define the Concept160Work with the user to nail down:161- What type of video (from categories above)?162- What's the subject?163- What's the start state and end state?164- What's the intended use (social, hero, course trailer)?165166### Step 2 — Generate Start/End Frame Prompts167If the user doesn't have start/end images yet, generate NB2 image prompts first:168- Use `/asset-generate` or `/asset-product-shot` for the start frame169- Use `/asset-generate` or `/asset-exploded-view` for the end frame170- Ensure both use identical lighting, camera angle, and palette171172### Step 3 — Write the Video Prompt173Compose the full 5-section video prompt. Show to user for approval.174175### Step 4 — Generate Variants176Offer 2-3 variants:1771. **Standard**: As described (e.g., 5 seconds, forward)1782. **Reverse**: End → Start (assembly instead of deconstruction)1793. **Loop**: Forward → Reverse seamlessly (for website backgrounds)180181### Step 5 — Deliver182Present all prompts in chat. If the user wants a polished delivery page, suggest `/asset-deliver`.183184## Output Format185186```187## Video Prompt Set: [CONCEPT NAME]188189### Concept190[1-2 sentence description of the video]191192### Use Case193[Where this will be used — social, hero section, course trailer, etc.]194195---196197### START FRAME PROMPT198[For image generation — paste into NB2/Midjourney/etc.]199200{start frame prompt}201202---203204### END FRAME PROMPT205[For image generation — paste into NB2/Midjourney/etc.]206207{end frame prompt}208209---210211### VIDEO TRANSITION PROMPT212[Paste into your video model, upload start frame as first frame, end frame as last frame]213214{full 5-section video prompt}215216---217218### VARIANT — Reverse Assembly219{reverse version}220221### VARIANT — Seamless Loop222{loop version}223224---225226### Recommended Settings227| Setting | Value |228|---------|-------|229| Duration | 5 seconds |230| Aspect | 16:9 |231| Frame rate | 24fps |232| Resolution | 1080p minimum |233| Model | Any (Runway, Kling, Pika, Higgsfield) |234235### Workflow2361. Generate start frame image with `/asset-generate` using the start frame prompt2372. Generate end frame image with `/asset-generate` using the end frame prompt2383. Upload both to your video model with the transition prompt2394. Use `/asset-brand-check` on the frames before generating video240```241242## Error Recovery243244| Issue | Fix |245|-------|-----|246| Video model produces jittery motion | Add "smooth, eased motion, no sudden jumps" and reduce complexity |247| Start/end frames look inconsistent | Regenerate end frame using start frame as reference in NB2 |248| Motion too fast | Increase duration, add "slow, deliberate, contemplative pace" |249| Background changes during transition | Add "background remains static and unchanged throughout" |250| Artifacts on moving elements | Simplify the transition — fewer moving parts, one clear axis of motion |251| Camera drifts | Add "locked-off tripod shot, absolutely no camera movement" |