Skip to content

AI generation: style option that references a second input image (facade reference) #1886

Description

@kfarr

Summary

Allow a second image input — a style / reference image — alongside the source render, so users can point at a specific building facade or material treatment instead of describing it in words.

Motivation

Paying customers ask for ability to more clearly specify design input for a given scene source image used in AI Render.

Proposed direction

Rather than a bare "second image" slot, fold this into the existing style selection:

  • The user picks a style, as today.
  • One style option is "reference image" (naming TBD), which reveals an image input.
  • The style prompt then refers to that second image — "match the facade treatment in image 2" — instead of hardcoding a look.

This keeps the UI at one concept ("what style?") rather than adding an unexplained second upload box, and reuses the mental model users already have.

Notes

  • Future / exploratory. No committed timeline.
  • Needs a check on which of the current image models accept multiple image inputs and how they reference them; the prompt phrasing is likely model-specific.
  • This relates to the request for single file picker via re-using / adapting the existing shared asset component, i forget which issue that is

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions