This workflow is based on my FLUX_2IMGto_1_Autoload-workflow. But now the second image is static and used as style template. For the first image you should create a directory with chosen images you want to use or just use an existing directory with images.

Updated to V1a with other Lora Stacker nodes and the new Gemini-node that is able to use Gemini 2.

V2 has a complete different Gemini node - LayerUtility: Gemini(Advance) - which substitutes the three I used before. So you make just one request instead of three. use just one prompt for Gemini instead of three. Further it allows you to use parameters as temperature, word limit, max tokens. Especially temperature has a big impact on the final prompt. Ialso addaed a third detailer - of course all detailer are optional - and placed the the scaler before the detailer.

V3: I used the new Gamini API by AbdallahAlswa and added Redux, In the pictures you can see the impact of the redux-option.