Video Mode

Try It Now

Video mode uses Google Veo 3.1 API to generate AI videos.

Basic Usage

1. Select "Video" mode
2. Choose generation method
3. Enter video description
4. Set video parameters
5. Click "Generate"

Generation Methods

Video mode offers four generation methods:

Text to Video

The most basic method - generate videos purely from text descriptions.

Frames to Video

Upload images as the video's start or end frames:

• Start Frame (required): The first frame of the video
• End Frame (optional): The last frame of the video
• Loop Video: If only uploading a start frame, you can create a looping video

References to Video

Upload reference images to guide video generation:

• Asset Reference: Make objects/characters in the video similar to the reference
• Style Reference: Make the video style similar to the reference
• Click the tag on the image to toggle reference type
• Maximum 3 reference images
• Prompt is required: You must enter a text description when using reference images

Limitations

When using reference images:

• Model is locked to Veo 3.1 (cannot use Fast model)
• Duration is locked to 8 seconds (cannot change)

Extend Video

Select a video from your history to generate extended content.

Limitations

• Only 720p resolution videos can be used as extension sources
• Each video can only be extended once (chain extension not supported)

Video Parameters

Model

• Veo 3.1 Fast: Faster generation, slightly lower quality
• Veo 3.1: High quality, requires more time

Video Length

• 4 seconds: Short clips, suitable for GIF-like use
• 6 seconds: Medium length
• 8 seconds: Standard length

Resolution

• 720p: Faster generation, smaller file size
• 1080p: High quality
• 4K: Ultra-high quality, requires more time

Aspect Ratio

• 16:9: Standard landscape
• 9:16: Portrait (mobile-friendly)

Prompt Tips

Describe Motion

Videos need dynamic scene descriptions:

A butterfly slowly takes off from a flower,
sunlight through the wings creates a rainbow halo,
background is a blurred garden

Cinematic Language

Using film terminology can achieve better results:

Slow motion close-up: water droplet falling from a leaf
Pan shot: path through the forest
Dolly shot: slowly approaching an ancient castle

Style Specification

Cinematic quality, shallow depth of field,
warm golden light,
35mm film grain

Negative Prompt

Video mode has a unique "Negative Prompt" feature that lets you specify elements you don't want in the video:

cartoon, painting, low quality, blurry, distorted

Audio Prompts

Veo 3.1 supports generating video audio. You can configure three types of sound in the "Audio Prompts" tab:

Dialogue

Make characters in the video speak specific content. Use quotes to indicate speech:

"This must be the key," he whispered

Tip

Use quotes for specific speech, and the model will attempt to have the character speak the quoted text.

Sound Effects

Describe specific sound effects in the video:

screeching tires, roaring engine

Ambient Sound

Set the background sound for the scene:

a faint eerie hum in the background

About Audio Generation

Currently, all generated videos will include audio (cannot be disabled).

This is because the generateAudio parameter is only supported by Vertex AI, not the Gemini API used by this application. Vertex AI requires server-side authentication (Service Account), which is incompatible with a fully client-side application.

If you need silent video functionality, you'll need to remove the audio track using external video editing software.

Generation Time

Video generation typically takes 1-3 minutes, depending on:

• Video length
• Resolution
• Content complexity
• API load

You can do other things during generation and will be notified when complete.

Cost Estimate

When configuring options, the interface displays a real-time cost estimate (price per second × video duration), helping you understand the cost before generating.

For actual pricing, see Google AI Pricing.

Download and Share

Completed videos can be:

• Downloaded as MP4 file
• Previewed and played
• Saved to history

Try It Now

Next Steps

• Image Generation - Generate key frames first
• History - Manage generated videos