Appearance
Video Generation Demo Script
Prerequisites
- GhostAI instance with Gemini API key configured
- Team, Business, or Enterprise tier account
- Optional: ffmpeg installed on server for storyboard export
Scene 1: Text-to-Video Generation (2 min)
Setup
Open a new GhostAI chat session.
Script
- Type prompt: "Generate a cinematic video of a sunset over a calm ocean with gentle waves and warm golden light"
- Watch the progress: Point out the inline progress indicator showing elapsed time and estimated completion
- Video appears: The 8-second video clip plays inline in the chat message
- Privacy angle: "Notice that the prompt was automatically scanned for PII before reaching the Veo API — the same tokenization pipeline that protects all your chat messages"
Talking Points
- Veo 3.1 generates photorealistic 8-second clips
- Same Gemini API key used for image generation — no additional configuration
- PII tokenization prevents accidental data leakage in prompts
- Videos are persisted to local storage + S3 for durability
Scene 2: Image-to-Video (Animate) (2 min)
Script
- Generate an image first: "Create an image of a mountain landscape at dawn with mist in the valley"
- Open artifact panel: Click on the generated image to view it in the artifact panel
- Click Animate: Hit the "Animate" button in the lightbox
- Observe: GhostAI automatically sends a prompt with the image URL to create a smooth video from the still image
- Compare: "The original image becomes the first frame, and the model adds natural camera movement and atmospheric effects"
Talking Points
- Zero-friction workflow from image to video
- One-click Animate button — no prompt engineering required
- Image-to-video uses the same PII protection as text-to-video
- Both image and video are stored and accessible in the conversation history
Scene 3: Multi-Clip Storyboard (3 min)
Script
- Generate 3 clips:
- "Generate a video of a sunrise timelapse over a city skyline"
- "Generate a video of people walking through a busy urban park"
- "Generate a video of a river flowing through a forested canyon"
- Open Storyboard Editor: Access from the artifact panel
- Demonstrate drag-to-reorder: Rearrange the clips on the timeline
- Preview: Click the Preview button to play the clips sequentially
- Export: Click "Export Video" to concatenate into a single file
- Download: Show the downloadable concatenated video
Talking Points
- Storyboard mode turns individual clips into coherent multi-scene videos
- Drag-and-drop reordering is instant — no round-trip to the server
- In-browser preview plays clips sequentially without server-side processing
- Export uses ffmpeg for seamless concatenation
- Each clip maintains its individual PII protection
Scene 4: Canvas App with Video (2 min)
Script
- Create a canvas: "Build me an interactive presentation with a hero video header, title overlay, and three feature cards below"
- Watch GhostAI: The model generates HTML that includes a
<video>tag referencing the previously generated video - Preview: The video plays directly in the canvas renderer
- Publish: Click Publish to make the canvas app public on GhostForge
- Visit published app: Open the public URL — the video plays without authentication
Talking Points
- Videos work seamlessly in GhostForge canvas apps
- No special configuration needed — the CSP already allows video media
- Published apps serve videos directly from the API without auth requirements
- This is unique to GhostAI — no other platform offers PII-safe video generation embedded in publishable apps
Tier & Quota Demo (1 min)
Script
- Show settings: Navigate to Settings > Subscription to show the video quota
- Point out tier table: Team gets 20/month, Business 100, Enterprise unlimited
- Mention SIEM: "Every video generation and quota event is logged to the SIEM —
video.generated,video.quota_exceeded"
Key Differentiators to Emphasize
- Privacy-first: Every video prompt is PII-tokenized — unique in the market
- Enterprise controls: Tier gating, monthly quotas, SIEM logging, audit trail
- Full workflow: Text → Image → Video → Storyboard → Published App
- Zero configuration: Uses the existing Gemini API key, no additional setup
- Open architecture: Works alongside 100+ LLM models, web search, code interpreter, and other tools