Is there a way to create images?

[croxarens]

Question

Trying the Gemini 2.5 Flesh Image model to create an image. The model replied that it generated the image, which I cannot see via the TUI, so I asked to save it, but nothing was saved.

Is there a way to let the model generate an image and save it?

[github-actions[bot]]

This issue might be a duplicate of or related to existing issues. Please check:

• #4951: imagen_generate tool crashes with validation error
• #6604: [FEATURE]: MCP Tool Output - image generation tool outputs not displayed in TUI
• #5362: Write Tool includes metadata in output affecting file content
• #7287: Model-generated output not reflected in TUI display
• #5006: TUI Renders Truncated LLM Response
• #4990: Output from agent is frequently hidden
• #6880: Subagents Output in TUI seems strange since 1.1.1
• #6951: Gemini 3 flash usage issue
• #6743: Google models list outdated - missing Gemini 2.5 Flash
• #7030: Tool calls show as executed but no files are created

The image creation issue appears to be caused by multiple factors: tool validation failures, TUI output visibility problems, and file persistence issues. Feel free to ignore if your specific case differs from these.

[ominiverdi]

I created a plugin that provides a workaround: opencode-antigravity-img

It uses the Gemini 3 Pro Image model via the Antigravity/CloudCode API and saves images directly to disk (bypassing TUI display issues).

Requires a Google AI subscription.

Setup:

"plugin": [
  "opencode-antigravity-auth",
  "opencode-antigravity-img"
]

Usage: "Generate an image of a cat jumping a fence"

Check quota: "Check my image quota" (resets every 5 hours)

Note: This doesn't fix the native TUI display issue - it's a plugin-based alternative.