Tool Usage: Text2Image ​

Overview ​

Text2Image refers to obtaining images through text descriptions.

Example: Completing Text2Image Using the Stable Diffusion Tool ​

Task ​

Use the text-to-image tool of stable diffusion to generate an image from a given text description.

Code ​

python examples/di/sd_tool_usage.py

python examples/di/sd_tool_usage.py

Execution Results ​

Produces a 512x768 image in metagpt's workspace/resources/SD_Output

Mechanism Explained ​

1. Standardize the existing code sd_engine.py under metagpt/tools to provide it to DataInterpreter. The tool definition is placed under metagpt/tools/functions/schemas, see stable_diffusion.yml.
2. When planning tasks, DataInterpreter automatically sets the intermediate task type as "stable_diffusion", scanning for available tools related to stable diffusion during task execution.
3. Upon discovering available tools: ['SDEngine'], it loads the tool code and automatically uses SD T2I (Text to Image) code to generate images according to current needs. Before generating images, it automatically expands the SD Prompt.