This guide covers the setup, usage, and best practices for image generation in elvex. It's organized using the Diataxis framework to cater to different user needs.
Step 1: Set up an Image Generation Provider
Before using image generation in any assistant, an Owner or Admin must configure an image generation model:
Go to Settings > Apps
Under Image generation provider, select a provider from the dropdown
Press the Save button
Step 2: Enable Image Generation for an Assistant
By default, assistants don't have image generation enabled. To enable it:
Go to your assistant settings screen
Expand the Tools section
Enable the the Image Generation tool
Step 3: Generate Your First Image
Once enabled, you can generate images by asking your assistant and providing a description:
Open a conversation with your assistant
Type a request like: "Create an image of a dog chasing a cat in a sunny, summer mountain landscape"
Specify the model you want to use (e.g., "using GPT Image 1" or "using DALL-E 3")
Note: If you encounter an error when using GPT Image 1, you may need to verify your OpenAI account. Follow the provided link to complete the verification process.
Crafting Effective Prompts
To get the best results from image generation:
Be specific in your descriptions
Include details about style, lighting, and composition
Use adjectives to convey mood and atmosphere
Example: "Generate an image of a futuristic cityscape at sunset, with flying cars and neon signs, in a cyberpunk style, using DALL-E 3"
Iterate on Generated Images
If the first result isn't quite what you wanted:
Analyze what aspects of the image you'd like to change
Modify your prompt to address these aspects
Try generating the image with a different model
Generate a new image with the updated prompt
Use Advanced Features
For GPT Image 1, you can create images with transparent backgrounds by specifying it in your prompt
Experiment with different image sizes and styles, which may vary depending on the model used
Image Generation Parameters
elvex supports multiple image generation models, each with its own capabilities:
Models
GPT Image 1
size: Options include "1024x1024" (default), "1536x1024" (landscape), "1024x1536" (portrait), or "auto"
background: Options are "transparent", "opaque", or "auto" (default)
DALL-E 3
size: Options are "1024x1024", "1792x1024", or "1024x1792"
style: Options are "vivid" (default, for hyper-real and dramatic images) or "natural" (for more natural-looking images)
DALL-E 2
size: Options are "256x256", "512x512", or "1024x1024"
Note: When using "auto" for size with GPT Image 1, it defaults to "1024x1024".
Conclusion
Image generation in elvex now offers more flexibility and power with support for multiple AI models. By understanding how to set up, use, and optimize your prompts across different models, you can harness the full potential of this advanced image generation capability. Remember to choose the appropriate model and parameters for your specific needs, and don't hesitate to experiment with different settings to achieve the best results.