Whole-Image Select + Conversational Prompt (Generative Fill Partner Models)

Overview

Instead of manually selecting each area to change before running Generative Fill, select the entire image (Cmd/Ctrl+A) and describe every intended edit — removals, additions, time-of-day or season shifts — in a single natural-language prompt. This is enabled by Photoshop's partner models, Google's Gemini 2.5 Nano Banana and Black Forest Labs' Flux Context Pro, which sit alongside Adobe's own Firefly Image Model 3 and Image Model 1 in the Generative Fill model picker and can be swapped from the Properties panel at any time without redoing the selection.

Why it works

Partner models appear to reason over a whole-image selection well enough to infer which regions to touch from prompt text alone, removing the precise manual selection/brushing step that Adobe's native Firefly models still benefit from. Compare to Generative Fill with Blank Prompt (Content-Aware Erase), which covers the older selection-driven pattern (brush a mask, leave the prompt blank) this technique is meant to replace for multi-part edits. Also related: Generative Fill Background Compositing Workflow for compositing use cases and Firefly Style Reference & Composition Reference for prompt-adjacent control mechanisms.

Practical notes

Partner models render one variation per click; Firefly Image Model 3 renders three variations per generation — factor this into credit spend.
Model quality is prompt- and scene-dependent, not uniformly won by one model: in one comparison, Nano Banana most often preserved original framing/scene fidelity, but Flux Context Pro won outright on the most complex multi-edit prompt and was sometimes judged more photorealistic. Running the identical prompt through multiple partner models before picking a result is the recommended practice, not a one-off demo trick.
Even the stronger model can leave artifacts on complex prompts (e.g., a first pass leaving some wires behind), requiring a second generation pass.
Each partner-model generation consumes paid credits regardless of whether the result is usable, and this cost isn't surfaced in the Photoshop UI itself.
Not appropriate for contexts with representational/liability stakes (e.g., removing permanent fixtures from real-estate listing photos) — treat as a creative/compositing technique, not a documentation-accuracy tool.

Из тем: Generative Fill & AI Compositing Toolkit