SciFig image models
Choose an image model for the scientific task in front of you
Model names matter less than the workflow. Compare the capabilities SciFig actually exposes, then open a model page with the generator already configured.
Models organized by practical use
Capabilities below are read from the same model registry used by the generator, so supported modes and limits stay aligned with the product.
GPT Image 2
Precise prompts, clean text
A flexible default when prompts, readable labels, image editing, and output-size control all matter.
Recommended for
Graphical abstracts, labeled mechanisms, and iterative edits
- Text to image
- Image to image
- 16 reference images
- Resolution controls up to 4K
Nano Banana family
3 variants
Three variants covering fast generation, multi-reference editing, and higher-quality image work.
Recommended for
Reference-guided revisions, rapid alternatives, and image-to-image work
- Text to image
- Image to image
- 14 reference images
- Resolution controls up to 4K
Z-Image
Photorealistic detail
A focused text-to-image option for photorealistic detail when editing and resolution controls are unnecessary.
Recommended for
Materials, instruments, environments, and cover concepts
- Text to image
- Image to image
- 0 reference images
- Fixed 1K output
Use model output as a visual draft, not scientific evidence
Every available model can invent labels, structures, quantities, and relationships. A qualified author must verify the result against the manuscript, source data, and target journal policy.