Imagen 3 is our current line of image generation models.
This page documents the capabilities and features of imagen-3.0-generate-002
.
Model ID
imagen-3.0-generate-002
Capabilities
- Supported
- Not supported
- Image customization using few-shot learning
- Subject customization for product, person, and animal companion
- Style customization
- Controlled customization
- Instruct customization or style transfer
- Mask-based image editing
- Mask-free image editing
- Insert objects in images
- Remove objects from images
- Outpainting
- Product image editing
- Upscale images
- Negative prompting
- Provisioned Throughput
Image ratios and resolutions
- 1:1: 1024x1024
- 3:4: 896x1280
- 4:3: 1280x896
- 9:16: 768x1408
- 16:9: 1408x768
Prompt languages
- English
- Chinese (simplified) Preview feature
- Chinese (traditional) Preview feature
- Hindi Preview feature
- Japanese Preview feature
- Korean Preview feature
- Portuguese Preview feature
- Spanish Preview feature
Limits
- Maximum API requests per minute per project: 20
- Maximum images returned per request (text-to-image generation): 4
- Maximum image size uploaded or sent in a request (MB): 10 MB
- Maximum input tokens (text-to-image generation prompt text): 480 tokens
For Imagen pricing information, see the Imagen section of the Cost of building and deploying AI models in Vertex AI page.
