PPT Generator Pro (NanoBanana PPT Skills)

Overview

Generate high-quality PPT slide images using Google Nano Banana Pro (Gemini 3 Pro Image Preview), with optional AI transition videos via Kling AI.

Tool location: /Users/carrick/tools_devop/NanoBanana-PPT-Skills/ Python venv: /Users/carrick/tools_devop/NanoBanana-PPT-Skills/venv/

Available Styles

styles/gradient-glass.md - Gradient frosted glass card style (tech/business)
styles/vector-illustration.md - Vector illustration style (education/creative)

Execution Workflow

Phase 1: Collect User Input

Document content: Read from file path or accept direct text
Style selection: Ask user to choose from available styles
Page count: Ask user (5 / 5-10 / 10-15 / 20-25 pages)
Resolution: 2K (2752x1536, recommended) or 4K (5504x3072)
Video: If Kling AI keys are configured, ask if transition videos are wanted

Phase 2: Content Planning

Analyze the document and create slides_plan.json:

{
  "title": "Document Title",
  "total_slides": 5,
  "slides": [
    {
      "slide_number": 1,
      "page_type": "cover",
      "content": "Title\nSubtitle"
    },
    {
      "slide_number": 2,
      "page_type": "content",
      "content": "Key points\n- Point 1\n- Point 2"
    }
  ]
}

Page types: cover, content, data

Save this file to: /Users/carrick/tools_devop/NanoBanana-PPT-Skills/slides_plan.json

Visual Continuity (Video Mode): If video transitions are requested, adjacent slides MUST share the same background style and color scheme. Kling AI's first-last frame feature requires similar images for smooth 5s transitions — large visual differences trigger hard camera cuts instead of smooth morphing. Use transition buffer pages between visually distinct sections if needed.

Phase 3: Generate PPT Images

cd /Users/carrick/tools_devop/NanoBanana-PPT-Skills && source venv/bin/activate && python3 generate_ppt.py \
  --plan slides_plan.json \
  --style styles/gradient-glass.md \
  --resolution 2K

Parameters:

--plan: Path to slides plan JSON
--style: Path to style template
--resolution: 2K or 4K
--output: Custom output directory (default: outputs/TIMESTAMP)

Phase 4: Generate Transition Prompts (Video Mode)

If video mode is selected, analyze generated PPT images and create transition_prompts.json. For each adjacent pair of slides:

Classify: Determine A-type (similar — same style/layout) or B-type (different — e.g. cover→content)
A-type strategy: "In-place evolution" — changes happen on subjects/environment, minimal camera movement. Prioritize: background gradient flow, 3D object rotation/morphing, card content fade in/out, lighting shifts
B-type strategy: "Camera-driven transition" — use explicit camera movement (push/pull/pan/rotate) as a bridge between different scenes

Kling AI first-last frame prompt rules:

Follow formula: subject + motion, background + motion
Keep prompts simple and concrete — describe what we SEE, not what we FEEL
All motion must be achievable within 5 seconds
Motion must follow physics — avoid complex physics (bouncing, throwing)
MUST end every prompt with text handling declaration (e.g. "text content transitions via fade, remaining clear and stable")
Avoid descriptions like "text changes/moves/rotates" — video models blur text

Refer to the full transition template at /Users/carrick/tools_devop/NanoBanana-PPT-Skills/prompts/transition_template.md for detailed creative framework.

Save prompts to the output directory as transition_prompts.json.

Phase 5: Generate Transition Videos (Video Mode)

cd /Users/carrick/tools_devop/NanoBanana-PPT-Skills && source venv/bin/activate && python3 generate_ppt_video.py \
  --slides-dir outputs/TIMESTAMP/images \
  --output-dir outputs/TIMESTAMP_video \
  --prompts-file outputs/TIMESTAMP/transition_prompts.json

Phase 6: Return Results

After generation, tell the user:

Output directory location
How to open the HTML viewer: open outputs/TIMESTAMP/index.html
Keyboard shortcuts: Arrow keys to navigate, Space for auto-play, ESC for fullscreen

Content Planning Guidelines

5 pages: Cover + 3 key points + Summary 5-10 pages: Cover + Introduction + 3-4 key points + Cases/Data + Summary 10-15 pages: Cover + TOC + 3 chapters (3 pages each) + Data visualization + Summary 20-25 pages: Cover + TOC + Introduction + 3 parts (4 pages each) + Case studies + Analysis + Summary

Kling AI Video Configuration

Model: kling-v2-6
Core feature: First-last frame (image-to-video) — upload start+end frame, generates transition
Mode: pro (high quality, 35 inspiration/5s) — first-last frame REQUIRES pro mode (std returns error 1201)
Duration: 5 seconds (recommended for PPT) or 10 seconds (×2 cost)
Aspect ratio: 16:9 (matches PPT images)
Prompt formula: subject + motion, background + motion

Inspiration cost estimate:

10-page PPT (9 transitions + 1 preview): 350 inspiration points (pro only)
Error 1102 Account balance not enough: Top up at https://klingai.kuaishou.com/

Performance

Image generation: ~30s/page (2K), ~60s/page (4K)
Transition video: ~30-60s/segment
Image size: ~2.5MB/page (2K), ~8MB/page (4K)

ppt-generator