NEW Browse AI tools across categories — updated daily. See what's new →

3d Animation Creator

Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation) and builds a production-quality website with scroll-driven animation. The video plays forward and ...

Version1.0.0
LicenseMIT
Token count~2,800
UpdatedJun 5, 2026

Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation) and builds a production-quality website with scroll-driven animation. The video plays forward and backward as the user scrolls, creating a mesmerizing Apple-style scroll-stopping effect. Uses frame extraction via FFmpeg, canvas-based rendering, and modern scroll-driven techniques. Includes: animated starscape background, annotation cards with snap-stop scroll, specs section with count-up animati...

Install

Quick install

via npx skills · works with 57+ agents
npx skills add https://github.com/sergeyramas/3d-animation-creator-skill
Or pick agent:
npx skills add sergeyramas/3d-animation-creator-skill --agent claude-code
npx skills add sergeyramas/3d-animation-creator-skill --agent cursor
npx skills add sergeyramas/3d-animation-creator-skill --agent codex
npx skills add sergeyramas/3d-animation-creator-skill --agent opencode
npx skills add sergeyramas/3d-animation-creator-skill --agent github-copilot
npx skills add sergeyramas/3d-animation-creator-skill --agent windsurf
More install options

Shorthand — useful for multi-skill repos:

npx skills add sergeyramas/3d-animation-creator-skill

Manual — clone the repo and drop the folder into your agent's skills directory:

git clone https://github.com/sergeyramas/3d-animation-creator-skill.git
cp -r 3d-animation-creator-skill ~/.claude/skills/
How to use: Once installed, ask your agent to "use the 3d-animation-creator skill" or describe what you want (e.g. "Takes a video file (e.g. a product deconstruction/assembly animation, before/aft"). Requires Node.js 18+.

3d-animation-creator

Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation) and builds a production-quality website with scroll-driven animation. The video plays forward and backward as the user scrolls, creating a mesmerizing Apple-style scroll-stopping effect. Uses frame extraction via FFmpeg, canvas-based rendering, and modern scroll-driven techniques. Includes: animated starscape background, annotation cards with snap-stop scroll, specs section with count-up animati...

---
name: 3d-animation-creator
description: >
Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation)
and builds a production-quality website with scroll-driven animation. The video plays forward and
backward as the user scrolls, creating a mesmerizing Apple-style scroll-stopping effect. Uses
frame extraction via FFmpeg, canvas-based rendering, and modern scroll-driven techniques. Includes:
animated starscape background, annotation cards with snap-stop scroll, specs section with count-up
animations, navbar with scroll-to-pill transform, loader, and full mobile responsiveness. Trigger
when the user says "3D animation", "scroll-stop build", "scroll animation website", "scroll-driven
video", "build the scroll-stop site", or provides a video file and asks to make it scroll-controlled.
Also trigger for "Apple-style scroll animation" or "video on scroll".
---

3D Animation Creator — Scroll-Driven Video Websites

You take a video file and build a production-quality website where the video playback is
controlled by scroll position — creating a dramatic, Apple-style scroll-stopping effect.

The user gives you a video. You handle everything: frame extraction, website build, content
population, and serving it locally for preview.

---

Step 0: The Interview (MANDATORY)

Before touching any code or extracting any frames, ask the user these questions.
Do not skip this step — the whole site is built from these answers.

Required Questions

Ask these naturally, not as a numbered interrogation:

  1. Brand name — "What's the brand or product name for this site?"
  2. Logo — "Do you have a logo file I can use? (SVG or PNG preferred)"
  3. Accent color — "What's your primary accent color? (hex code, or describe it and I'll suggest options)"
  4. Background color — "What background color? (dark backgrounds work best for this effect)"
  5. Overall vibe — "What feel are you going for? (e.g., premium tech launch, luxury, playful, minimal, bold)"

Content Sourcing

Ask how they want to provide the website content:

  • Option A: From an existing website — "Share the URL and I'll pull the real content (product name, features, specs, copy)."
  • Option B: Paste it in — "Paste product descriptions, feature lists, specs, testimonials — whatever you have."

If they provide a URL, use WebFetch to retrieve the page and extract relevant copy, product details, features, specs, and any other usable content.

Optional Sections

Ask whether they want these included:

  • Testimonials — "Want a testimonials section? Provide them or I'll pull from the URL you shared."
  • Confetti — "Want a confetti burst effect anywhere? (e.g., on CTA button click)"
  • Card Scanner — "Want a 3D particle showcase section? (Three.js-based — good for showing off a card, device, or object)"

Only include these if the user explicitly opts in.

---

Prerequisites

  • FFmpeg must be installed (brew install ffmpeg if not)
  • The user provides a video file (MP4, MOV, WebM, etc.)
  • Video should be relatively short (3-10 seconds is ideal)
  • The first frame of the video MUST be on a white background. If it isn't, let the user know and ask for a re-export or a separate white-background hero image.

---

Design System (Built from User's Answers)

Once the interview is complete, construct the design system:

  • Fonts: Space Grotesk (headings), Archivo (body), JetBrains Mono (code/mono)
  • Accent color: From user's answer — used for buttons, glows, progress bars, highlights
  • Background color: From user's answer — used for body, sections
  • Text colors: Derived from background — dark bg gets white primary + muted secondary; light bg gets dark primary + muted secondary
  • Selection: Accent color background with contrasting text
  • Scrollbar: Dark track with gradient thumb using accent color, glow on hover
  • Cards: Glass-morphism — semi-transparent bg, subtle border, backdrop-filter: blur(20px), border-radius: 20px
  • Buttons: Primary = accent color bg with contrasting text + accent glow; Secondary = transparent with border
  • Effects: Floating background orbs (accent color tones, blurred), subtle grid overlay, animated starscape
  • Brand name & logo: Used in navbar, footer, loader, and anywhere branding appears

---

The Build Process

Step 1: Analyze the Video

ffprobe -v quiet -print_format json -show_streams -show_format "{VIDEO_PATH}"

Extract duration, fps, resolution, total frame count. Target 60-150 frames total.

Step 2: Extract Frames

mkdir -p "{OUTPUT_DIR}/frames"
ffmpeg -i "{VIDEO_PATH}" -vf "fps={TARGET_FPS},scale=1920:-2" -q:v 2 "{OUTPUT_DIR}/frames/frame_%04d.jpg"

Use -q:v 2 for high quality JPEG. Use JPEG not PNG for smaller files.

Step 3: Build the Website

Create a single HTML file. The site has these sections (top to bottom):

  1. Starscape — Fixed canvas behind everything with ~180 twinkling animated stars
  2. Loader — Full-screen with brand logo, "Loading" text, accent-colored progress bar
  3. Scroll Progress Bar — Fixed top, accent gradient, 3px tall
  4. Navbar — Brand logo + name, transforms from full-width to centered pill on scroll
  5. Hero — Title, subtitle, CTA buttons, scroll hint, background orbs + grid
  6. Scroll Animation — Sticky canvas with frame sequence, annotation cards with snap-stop
  7. Specs — Four stat numbers with count-up animation on scroll
  8. Features — Glass-morphism cards in a grid
  9. CTA — Call to action section
  10. Testimonials(only if user opted in) Horizontal drag-to-scroll testimonial cards
  11. Card Scanner(only if user opted in) Three.js particle showcase
  12. Footer — Brand name and links

For the exact CSS/JS implementation of every section, read references/sections-guide.md.
That file contains the complete code for each section — structure, styling, JavaScript, mobile
adaptations, and animation patterns.

Step 4: Key Implementation Patterns

Canvas rendering with Retina support:

canvas.width = window.innerWidth * window.devicePixelRatio;
canvas.height = window.innerHeight * window.devicePixelRatio;
canvas.style.width = window.innerWidth + 'px';
canvas.style.height = window.innerHeight + 'px';

Cover-fit drawing (desktop) — zoomed contain-fit (mobile):
On desktop, use cover-fit so the frame fills edge-to-edge. On mobile, use a slightly zoomed
contain-fit approach so the object stays centered and visible.

Annotation cards with snap-stop scroll:
Annotation cards appear at specific scroll progress points (data-show/data-hide attributes).
The scroll FREEZES briefly at each card position — creating a "boom, boom, boom" effect where
each card pops up as you stop. Uses JS-based snap: detects when scroll progress enters a snap zone,
scrolls to the exact position, locks the body overflow for ~600ms, then releases.
The number of annotation cards is flexible — match it to the content the user provides.

Navbar scroll-to-pill transform:
The navbar starts full-width, then on scroll shrinks to a centered pill shape (max-width ~820px)
with rounded corners and glass-morphism background.

Count-up animation:
Spec numbers animate from 0 to target with easeOutExpo easing, staggered 200ms apart.
Numbers get an accent-color glow pulse while counting. Triggered by IntersectionObserver.

Animated starscape:
A fixed canvas behind everything with ~180 stars that slowly drift and twinkle. Each star has
random drift speed, twinkle speed/phase, and opacity. Creates a subtle living background.

Step 5: Customize Content

All content comes from the interview (Step 0). Use the real brand name, real product details,
and real copy — never use placeholder "Lorem ipsum" text. If content came from a website URL,
use the actual text from that site. Populate:

  • Hero title and subtitle
  • Annotation card labels, descriptions, and stats
  • Spec numbers and labels
  • Feature cards
  • CTA text
  • Testimonials (if included)

Step 6: Serve & Test

cd "{OUTPUT_DIR}" && python3 -m http.server 8080

Open http://localhost:8080 and test. Then open the browser URL for the user.

---

Mobile Responsiveness

Key mobile adaptations:

  • Annotation cards: Compact single-line design — hide paragraph text, stat numbers, and labels. Show only card number + title in a flex row. Position at bottom of viewport (bottom: 1.5vh)
  • Scroll animation height: 350vh desktop, 300vh tablet, 250vh phone
  • Navbar: Hide links on mobile, show only logo + pill shape
  • Testimonials (if included): Touch-scrollable, snap to card edges
  • Feature cards: Stack to single column
  • Specs: 2x2 grid on mobile

---

Best Practices

  1. requestAnimationFrame for drawing — Never draw directly in scroll handler
  2. { passive: true } on scroll listener — Enables browser scroll optimizations
  3. Canvas with devicePixelRatio — Crisp on Retina displays
  4. Preload all frames before showing — No pop-in during scroll
  5. Frame deduplication — Only call drawFrame when frame index actually changes
  6. No scroll-behavior: smooth — Would interfere with frame-accurate scroll mapping
  7. No heavy JS libraries — Pure vanilla JS except Three.js for card scanner (if included)
  8. Sticky canvasposition: sticky keeps canvas viewport-fixed while scroll container moves
  9. White first frame — The video must start on a clean white background

---

Error Recovery

| Issue | Solution |
|---|---|
| FFmpeg not installed | Tell user to run brew install ffmpeg |
| Frames don't load | Check file paths, ensure local server is running (can't load from file://) |
| Animation is choppy | Reduce frame count, ensure JPEG not PNG, check file sizes (<100KB each) |
| Canvas is blurry | Ensure devicePixelRatio scaling is applied |
| Scroll feels too fast/slow | Adjust .scroll-animation height (200vh=fast, 500vh=slow, 800vh=cinematic) |
| Mobile cards overlap content | Use compact single-line card design, position at bottom: 1.5vh |
| Snap-stop feels jarring | Reduce HOLD_DURATION to 400ms or increase SNAP_ZONE |
| Stars too bright/dim | Adjust starscape canvas opacity (default 0.6) |
| First frame isn't white | Ask user to re-export video with white opening frame |
| Video too long (>10s) | Recommend trimming to 3-6 seconds for optimal scroll feel |

---

Source: https://github.com/sergeyramas/3d-animation-creator-skill
Author: sergeyramas
Discovered via: skillsdirectory.com
Genre: ai-agents

SKILL.md source

---
name: 3d-animation-creator
description: Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation) and builds a production-quality website with scroll-driven animation. The video plays forward and ...
---

# 3d-animation-creator

Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation) and builds a production-quality website with scroll-driven animation. The video plays forward and backward as the user scrolls, creating a mesmerizing Apple-style scroll-stopping effect. Uses frame extraction via FFmpeg, canvas-based rendering, and modern scroll-driven techniques. Includes: animated starscape background, annotation cards with snap-stop scroll, specs section with count-up animati...

---
name: 3d-animation-creator
description: >
  Takes a video file (e.g. a product deconstruction/assembly animation, before/after transformation)
  and builds a production-quality website with scroll-driven animation. The video plays forward and
  backward as the user scrolls, creating a mesmerizing Apple-style scroll-stopping effect. Uses
  frame extraction via FFmpeg, canvas-based rendering, and modern scroll-driven techniques. Includes:
  animated starscape background, annotation cards with snap-stop scroll, specs section with count-up
  animations, navbar with scroll-to-pill transform, loader, and full mobile responsiveness. Trigger
  when the user says "3D animation", "scroll-stop build", "scroll animation website", "scroll-driven
  video", "build the scroll-stop site", or provides a video file and asks to make it scroll-controlled.
  Also trigger for "Apple-style scroll animation" or "video on scroll".
---

# 3D Animation Creator — Scroll-Driven Video Websites

You take a video file and build a production-quality website where the video playback is
controlled by scroll position — creating a dramatic, Apple-style scroll-stopping effect.

The user gives you a video. You handle everything: frame extraction, website build, content
population, and serving it locally for preview.

---

## Step 0: The Interview (MANDATORY)

Before touching any code or extracting any frames, ask the user these questions.
Do not skip this step — the whole site is built from these answers.

### Required Questions

Ask these naturally, not as a numbered interrogation:

1. **Brand name** — "What's the brand or product name for this site?"
2. **Logo** — "Do you have a logo file I can use? (SVG or PNG preferred)"
3. **Accent color** — "What's your primary accent color? (hex code, or describe it and I'll suggest options)"
4. **Background color** — "What background color? (dark backgrounds work best for this effect)"
5. **Overall vibe** — "What feel are you going for? (e.g., premium tech launch, luxury, playful, minimal, bold)"

### Content Sourcing

Ask how they want to provide the website content:

- **Option A: From an existing website** — "Share the URL and I'll pull the real content (product name, features, specs, copy)."
- **Option B: Paste it in** — "Paste product descriptions, feature lists, specs, testimonials — whatever you have."

If they provide a URL, use `WebFetch` to retrieve the page and extract relevant copy, product details, features, specs, and any other usable content.

### Optional Sections

Ask whether they want these included:

- **Testimonials** — "Want a testimonials section? Provide them or I'll pull from the URL you shared."
- **Confetti** — "Want a confetti burst effect anywhere? (e.g., on CTA button click)"
- **Card Scanner** — "Want a 3D particle showcase section? (Three.js-based — good for showing off a card, device, or object)"

Only include these if the user explicitly opts in.

---

## Prerequisites

- **FFmpeg** must be installed (`brew install ffmpeg` if not)
- The user provides a video file (MP4, MOV, WebM, etc.)
- Video should be relatively short (3-10 seconds is ideal)
- **The first frame of the video MUST be on a white background.** If it isn't, let the user know and ask for a re-export or a separate white-background hero image.

---

## Design System (Built from User's Answers)

Once the interview is complete, construct the design system:

- **Fonts**: Space Grotesk (headings), Archivo (body), JetBrains Mono (code/mono)
- **Accent color**: From user's answer — used for buttons, glows, progress bars, highlights
- **Background color**: From user's answer — used for body, sections
- **Text colors**: Derived from background — dark bg gets white primary + muted secondary; light bg gets dark primary + muted secondary
- **Selection**: Accent color background with contrasting text
- **Scrollbar**: Dark track with gradient thumb using accent color, glow on hover
- **Cards**: Glass-morphism — semi-transparent bg, subtle border, `backdrop-filter: blur(20px)`, `border-radius: 20px`
- **Buttons**: Primary = accent color bg with contrasting text + accent glow; Secondary = transparent with border
- **Effects**: Floating background orbs (accent color tones, blurred), subtle grid overlay, animated starscape
- **Brand name & logo**: Used in navbar, footer, loader, and anywhere branding appears

---

## The Build Process

### Step 1: Analyze the Video

```bash
ffprobe -v quiet -print_format json -show_streams -show_format "{VIDEO_PATH}"
```

Extract duration, fps, resolution, total frame count. Target 60-150 frames total.

### Step 2: Extract Frames

```bash
mkdir -p "{OUTPUT_DIR}/frames"
ffmpeg -i "{VIDEO_PATH}" -vf "fps={TARGET_FPS},scale=1920:-2" -q:v 2 "{OUTPUT_DIR}/frames/frame_%04d.jpg"
```

Use `-q:v 2` for high quality JPEG. Use JPEG not PNG for smaller files.

### Step 3: Build the Website

Create a single HTML file. The site has these sections (top to bottom):

1. **Starscape** — Fixed canvas behind everything with ~180 twinkling animated stars
2. **Loader** — Full-screen with brand logo, "Loading" text, accent-colored progress bar
3. **Scroll Progress Bar** — Fixed top, accent gradient, 3px tall
4. **Navbar** — Brand logo + name, transforms from full-width to centered pill on scroll
5. **Hero** — Title, subtitle, CTA buttons, scroll hint, background orbs + grid
6. **Scroll Animation** — Sticky canvas with frame sequence, annotation cards with snap-stop
7. **Specs** — Four stat numbers with count-up animation on scroll
8. **Features** — Glass-morphism cards in a grid
9. **CTA** — Call to action section
10. **Testimonials** — *(only if user opted in)* Horizontal drag-to-scroll testimonial cards
11. **Card Scanner** — *(only if user opted in)* Three.js particle showcase
12. **Footer** — Brand name and links

**For the exact CSS/JS implementation of every section**, read `references/sections-guide.md`.
That file contains the complete code for each section — structure, styling, JavaScript, mobile
adaptations, and animation patterns.

### Step 4: Key Implementation Patterns

**Canvas rendering with Retina support:**
```javascript
canvas.width = window.innerWidth * window.devicePixelRatio;
canvas.height = window.innerHeight * window.devicePixelRatio;
canvas.style.width = window.innerWidth + 'px';
canvas.style.height = window.innerHeight + 'px';
```

**Cover-fit drawing (desktop) — zoomed contain-fit (mobile):**
On desktop, use cover-fit so the frame fills edge-to-edge. On mobile, use a slightly zoomed
contain-fit approach so the object stays centered and visible.

**Annotation cards with snap-stop scroll:**
Annotation cards appear at specific scroll progress points (data-show/data-hide attributes).
The scroll FREEZES briefly at each card position — creating a "boom, boom, boom" effect where
each card pops up as you stop. Uses JS-based snap: detects when scroll progress enters a snap zone,
scrolls to the exact position, locks the body overflow for ~600ms, then releases.
The number of annotation cards is flexible — match it to the content the user provides.

**Navbar scroll-to-pill transform:**
The navbar starts full-width, then on scroll shrinks to a centered pill shape (max-width ~820px)
with rounded corners and glass-morphism background.

**Count-up animation:**
Spec numbers animate from 0 to target with easeOutExpo easing, staggered 200ms apart.
Numbers get an accent-color glow pulse while counting. Triggered by IntersectionObserver.

**Animated starscape:**
A fixed canvas behind everything with ~180 stars that slowly drift and twinkle. Each star has
random drift speed, twinkle speed/phase, and opacity. Creates a subtle living background.

### Step 5: Customize Content

All content comes from the interview (Step 0). Use the real brand name, real product details,
and real copy — never use placeholder "Lorem ipsum" text. If content came from a website URL,
use the actual text from that site. Populate:

- Hero title and subtitle
- Annotation card labels, descriptions, and stats
- Spec numbers and labels
- Feature cards
- CTA text
- Testimonials (if included)

### Step 6: Serve & Test

```bash
cd "{OUTPUT_DIR}" && python3 -m http.server 8080
```

Open `http://localhost:8080` and test. Then open the browser URL for the user.

---

## Mobile Responsiveness

Key mobile adaptations:

- **Annotation cards**: Compact single-line design — hide paragraph text, stat numbers, and labels. Show only card number + title in a flex row. Position at bottom of viewport (`bottom: 1.5vh`)
- **Scroll animation height**: `350vh` desktop, `300vh` tablet, `250vh` phone
- **Navbar**: Hide links on mobile, show only logo + pill shape
- **Testimonials** (if included): Touch-scrollable, snap to card edges
- **Feature cards**: Stack to single column
- **Specs**: 2x2 grid on mobile

---

## Best Practices

1. **`requestAnimationFrame` for drawing** — Never draw directly in scroll handler
2. **`{ passive: true }` on scroll listener** — Enables browser scroll optimizations
3. **Canvas with `devicePixelRatio`** — Crisp on Retina displays
4. **Preload all frames before showing** — No pop-in during scroll
5. **Frame deduplication** — Only call `drawFrame` when frame index actually changes
6. **No `scroll-behavior: smooth`** — Would interfere with frame-accurate scroll mapping
7. **No heavy JS libraries** — Pure vanilla JS except Three.js for card scanner (if included)
8. **Sticky canvas** — `position: sticky` keeps canvas viewport-fixed while scroll container moves
9. **White first frame** — The video must start on a clean white background

---

## Error Recovery

| Issue | Solution |
|---|---|
| FFmpeg not installed | Tell user to run `brew install ffmpeg` |
| Frames don't load | Check file paths, ensure local server is running (can't load from `file://`) |
| Animation is choppy | Reduce frame count, ensure JPEG not PNG, check file sizes (<100KB each) |
| Canvas is blurry | Ensure `devicePixelRatio` scaling is applied |
| Scroll feels too fast/slow | Adjust `.scroll-animation` height (200vh=fast, 500vh=slow, 800vh=cinematic) |
| Mobile cards overlap content | Use compact single-line card design, position at `bottom: 1.5vh` |
| Snap-stop feels jarring | Reduce HOLD_DURATION to 400ms or increase SNAP_ZONE |
| Stars too bright/dim | Adjust starscape canvas opacity (default 0.6) |
| First frame isn't white | Ask user to re-export video with white opening frame |
| Video too long (>10s) | Recommend trimming to 3-6 seconds for optimal scroll feel |


---

**Source**: https://github.com/sergeyramas/3d-animation-creator-skill
**Author**: sergeyramas
**Discovered via**: skillsdirectory.com
**Genre**: ai-agents

Related skills 6

running-claude-code-via-litellm-copilot

★ Featured

Use when routing Claude Code through a local LiteLLM proxy to GitHub Copilot, reducing direct Anthropic spend, configuring ANTHROPIC_BASE_URL or ANTHROPIC_MODEL overrides, or troubleshooting Copilot proxy setup failures such as model-not-found, no localhost traffic, or GitHub 401/403 auth errors.

xixu-me 155k
AI & ML

skills-cli

★ Featured

Use when users ask to discover, install, list, check, update, remove, back up, restore, sync, or initialize Agent Skills, mention `bunx skills`, `npx skills`, `skills.sh`, or `skills-lock.json`, ask "find a skill for X", or want help extending agent capabilities with installable skills.

xixu-me 155k
AI & ML

repo-intake-and-plan

★ Featured

Narrow RigorPilot helper for README-first deep learning repo reproduction. Use when the task is specifically to scan a repository, read the README and common project files, extract documented commands, classify inference, evaluation, and training candidates, and return the smallest trustworthy reproduction plan to the main orchestrator. Do not use for environment setup, asset download, command execution, final reporting, paper lookup, or end-to-end orchestration.

lllllllama 127k
AI & ML

image-to-video

★ Featured

Animate any still image on RunComfy — this skill is a smart router that matches the user's intent to the right i2v model in the RunComfy catalog. Picks HappyHorse 1.0 I2V (Arena #1, native audio, identity preservation) for general animations, Wan 2.7 with `audio_url` for custom-voiceover lip-sync, or Seedance 2.0 Pro for multi-modal animation from image + reference video + reference audio. Bundles each model's documented prompting patterns so the caller gets sharper output without burning ite...

agentspace-so 121k
AI & ML

video-edit

★ Featured

Edit existing video on RunComfy — this skill is a smart router that matches the user's intent to the right edit model in the RunComfy catalog. Picks Wan 2.7 Edit-Video (general restyle / background swap / packaging swap, identity + motion preservation), Kling 2.6 Pro Motion Control (transfer precise motion from a reference video to a target character), or Lucy Edit Restyle (lightweight identity-stable restyle / outfit swap). Bundles each model's documented prompting patterns so the skill gets...

agentspace-so 121k
AI & ML

nano-banana-2

★ Featured

Generate images with Google Nano Banana 2 (Gemini-family flash-tier text-to-image) on RunComfy — bundled with the model's documented prompting patterns so the skill gets sharper output than naive prompting against the same model. Documents Nano Banana 2's strengths (rapid iteration, in-image typography rendering, predictable framing, optional web-grounded context), the resolution-tier pricing, the safety-tolerance dial, and when to route to Nano Banana Pro / GPT Image 2 / Flux 2 / Seedream in...

agentspace-so 121k
AI & ML