Skip to main content

Supported Text-to-Image Models

Ideon currently supports these model IDs:

  • black-forest-labs/flux-schnell (default)
  • black-forest-labs/flux-2-pro
  • bytedance/seedream-4
  • google/nano-banana-pro
  • prunaai/z-image-turbo

Override Handling

User-provided t2i.inputOverrides are sanitized before runtime:

  • Non-user-configurable fields are dropped
  • Values are coerced and validated by type
  • Invalid values are ignored

Pipeline-Managed Fields

Depending on model capabilities, Ideon may set pipeline fields automatically:

  • aspect_ratio
  • width
  • height

This preserves stable image sizing and layout intent.

Selection Tips

  • Start with default flux-schnell for speed
  • Increase quality-focused fields only after baseline stability
  • Use dry-run to validate orchestration before live image spend

Practical Tradeoffs

ModelSpeedTypical QualityBest Use
black-forest-labs/flux-schnellFastGood baselineIteration, high-throughput drafts
black-forest-labs/flux-2-proMediumHigher fidelityHero visuals, polished cover images
bytedance/seedream-4MediumStrong compositionBalanced speed/quality workflows
google/nano-banana-proMediumStyle-consistent outputsBrand-like variation and consistency
prunaai/z-image-turboFastLightweightQuick exploratory rendering

Common Override Patterns

Use t2i.inputOverrides when you need model-specific tuning:

  • Output format tuning (for example png vs webp)
  • Model-native quality knobs where supported
  • Count/variant controls where allowed

Example:

{
"settings": {
"t2i": {
"modelId": "black-forest-labs/flux-schnell",
"inputOverrides": {
"output_format": "png"
}
}
}
}

If an override is invalid for the selected model, Ideon drops it during validation.