Google Gemini / Veo

Google Gemini is a multi-modal AI assistant that supports text, image (Imagen), and video (Veo) generation. Veo 3 is Google's latest video generation model, capable of producing high-quality video clips with advanced camera control and consistency.

Platform Overview

• Provider: Google DeepMind
• Specialty: Multi-modal (text, image, video)
• Access: Google account required, some features require subscription
• Supported Output Types: Video, Image

Lazer Detection

URL Patterns

The extension detects Gemini/Veo on these domains:

• https://gemini.google.com
• https://aistudio.google.com
• https://deepmind.google.* (with "veo" in URL)
• https://labs.google.* (with "video" in URL)

What Is Auto-Detected

When you generate with Gemini or Veo, Lazer extracts:

• Prompt - From contenteditable chat input or textarea
• Model - From model selector dropdown (e.g., "Gemini 2.5 Pro", "Veo 3")
• Temperature - From AI Studio advanced settings
• Aspect Ratio - From ratio selector buttons
• Duration - From duration slider (Veo video)
• Output URL - Video/image src from generation results
• Thumbnail - Video poster or image preview

Detection Quality

• Prompt extraction: High - Works in chat and AI Studio
• Settings extraction: High - Most settings visible in UI
• Output extraction: High - Reliable media URL detection

Capture Workflow

Gemini Chat Interface

1. Open Gemini chat at gemini.google.com
2. Type your prompt (e.g., "Generate an image of a mountain sunset")
3. Send message
4. Wait for Gemini to generate image/video
5. Open Lazer extension side panel
6. Detection banner shows "Google Gemini / Veo"
7. Click "Auto Fill"
8. Review and adjust fields
9. Click "Save Capture"

AI Studio Interface

1. Open AI Studio at aistudio.google.com
2. Choose "Imagen" or "Veo" from model selector
3. Enter prompt in prompt editor
4. Adjust settings (aspect ratio, duration, temperature)
5. Click "Run"
6. Wait for generation to complete
7. Open Lazer extension
8. Click "Auto Fill"
9. Click "Save Capture"

Multiple Outputs

Gemini sometimes generates multiple variants:

1. Capture first variant as usual
2. Scroll to next variant
3. Click "Refresh Detection" in extension
4. Click "Auto Fill" again
5. Capture each variant individually

Capture Tips

Best Practices

• Use AI Studio for control - More settings exposed than chat interface
• Note watermarking - Veo adds SynthID watermark (invisible provenance)
• Check model version - Model selector shows exact version (e.g., "Veo 3")
• Wait for load - Images/videos load asynchronously, wait for completion

SynthID Watermarking

Google embeds SynthID watermarks in Veo-generated videos:

• Invisible to human eye
• Machine-detectable provenance marker
• Cannot be removed without degrading video quality
• Lazer captures this in metadata as synthIDPresent: true

Common Issues

Prompt extracted from wrong message:

• Gemini chat has many messages
• Detector uses last user message
• Verify prompt text in capture form before saving

Settings not detected in chat:

• Chat interface hides advanced settings
• Use AI Studio for full settings control
• Manually fill settings fields if needed

Multiple images in one response:

• Gemini may generate 2-4 image variants
• Capture each separately by refreshing detection

Rights and Licensing

Ownership

Per Google's terms:

• Gemini outputs - You retain ownership of your outputs
• Commercial use - Allowed for most content types
• Attribution - Not required for most uses

SynthID and Provenance

• Veo videos include SynthID watermark
• Helps verify authenticity and origin
• Supports C2PA provenance standards (future)

Restrictions

• Subject to Google's Prohibited Use Policy
• Cannot use for illegal or harmful content
• Some beta features may have additional restrictions

Lazer Rights State

The extension sets rightsState to:

• LICENSED - Default for authenticated Gemini users
• UNKNOWN - If plan status unclear

Manually update in web app if needed.

Known Limitations

UI Variations

Gemini has multiple interfaces:

• Chat (gemini.google.com)
• AI Studio (aistudio.google.com)
• Integrated into Google Workspace

Detector supports chat and AI Studio primarily.

Async Loading

Images and videos load asynchronously:

• May take 5-10 seconds after generation completes
• Click "Refresh Detection" if output not found
• Extension detects when DOM is updated

Model Naming

Model selector sometimes shows:

• "Gemini 2.5 Pro" (text model)
• "Imagen 3" (image model)
• "Veo 3" (video model)

Detector captures exact model name displayed.

Metadata Fields

Gemini/Veo-specific metadata:

{
  "modelName": "Veo 3",
  "aspectRatio": "16:9",
  "duration": 8,
  "temperature": 0.7,
  "synthIDPresent": true,
  "platform": "google-veo"
}

Example Capture

A typical Veo capture in Lazer:

• Platform: google-veo
• Asset Type: VIDEO
• Prompt: "A drone flyover of a misty forest at dawn, cinematic lighting"
• Model: Veo 3
• Output URL: https://storage.googleapis.com/video_xyz789.mp4
• Thumbnail: https://storage.googleapis.com/thumb_xyz789.jpg
• Metadata: {"duration": 8, "aspectRatio": "16:9", "synthIDPresent": true}
• Status: GENERATED
• Rights State: LICENSED

Troubleshooting

Detection Not Working

• Verify you're logged into Google account
• Check that generation has completed
• Refresh detection after content loads
• Grant host permissions for gemini.google.com and aistudio.google.com

Output URL Expired

• Google uses googleusercontent.com and storage.googleapis.com CDNs
• URLs may be temporary
• Extension syncs immediately
• Verify sync in Queue mode

Model Not Detected

• AI Studio shows model clearly
• Chat interface may not display model badge
• Manually fill model field if needed

SynthID Metadata

To verify SynthID presence:

• Use Google's SynthID detector tool (future)
• Metadata field captures synthIDPresent: true when detected
• Currently inferred, not directly verified

Related Resources

• Google Gemini
• Google AI Studio
• Veo Video Generation
• SynthID Watermarking

Next Steps

• Compare with OpenAI Sora
• Understanding provenance tracking
• Capture workflow guide