PhotoGPTGuides

PhotoGPT Advanced Settings for Image Generation

In this guide, we'll go through the advanced options available in your dashboard for higher control over your generated images.

Written by PhotoGPT AI

Pre-requisites before you dive into advanced settings

Before going through this guide, please ensure that you are familiar with the basic image generation and model training steps. If you want to revisit and freshen up your memory, kindly refer to the below guides for context:

Assuming you are confident in the basics, let's talk about some advanced options and how you can use them to get higer control over your generations and achieve even better results.

Advanced Settings and their Utility

There are two kinds of image generations possible within the PhotoGPT Dashboard:

  • The first is character-based generation, where you can select one of your trained models or a built-in character model. Built-in models do not require personal-model training; generation still uses credits.
  • The second one is the general purpose image generation for any use case. This kind does not require any training and can be used for a single click generation. These are specialized models like Nano Banana and Seedream that can be used for varied purpose.

Below we'll talk about the advanced settings available within both of these type of generations and how they can be used for achieving best suited results for your use case.

1. Character-based Generation

Go to your image generation dashboard, select a character model you want to work with from the model drop-down in the left pane and then click on "Advance Settings". This will open up some additional settings as shown in the below screenshots. Listed below are a short description of what each of these options are and how you can leverage them for achieving results suited better to your needs:

  1. Steps: Choose how many generation steps to use. Character models default to 30 and accept 20–100 in increments of 5. Start with the default, then compare results before increasing the setting; more steps do not guarantee better prompt adherence.
  2. Seed: Fix the seed while keeping the model, prompt, and other settings unchanged to make variations easier to compare. Image details show the recorded seed where available.
  3. Resolution: Choose a pixel-size preset that suits your composition, such as 896 × 896 for a square image, 768 × 1344 for a portrait, or 1344 × 768 for a landscape.
  4. Reference Image Style Strength: Adjust the reference image's influence from 0 to 1. The default is 0.3; compare results to find a balance between the reference and your prompt.
  5. Enhance Face: Enable this option to request face enhancement with a character model. Review the result for facial detail and likeness; completion time varies.

Beyond these advanced settings, once you generate an image, you can use the image upscaler to get higher resolution or the AI Editor for a finishing touch.

2. General-purpose Generation

Select a general-purpose model from the same dropdown. These eight models expose image-size and aspect-ratio choices, with additional controls depending on the provider:

ModelImage sizesAdditional controls
GPT Image 2.51K, 2K, 4KQuality: low, medium, high, xhigh, max
GPT Image 21K, 2K, 4KQuality: low, medium, high
Gemini Nano Banana1KSafety Filter, Sequential Mode
Gemini Nano Banana Pro1K, 2K, 4KSafety Filter, Sequential Mode
Gemini Nano Banana 2.11K, 2K, 4KSafety Filter, Sequential Mode
Seedream 4.01K, 2K, 4KOptimize Mode, Sequential Mode
Seedream 4.52K, 4KOptimize Mode, Sequential Mode
Seedream 5.0 Lite2K, 3KOptimize Mode, Sequential Mode
  • Image Size and Aspect Ratio: Choose the size offered by your model and the composition you need, such as 1:1, 9:16, or 16:9. Size and quality can change the credit cost.
  • Quality: GPT Image models offer the quality levels listed above. Review the credit estimate when changing this setting.
  • Safety Filter: Gemini models expose a filter-level choice. The selected model and platform content rules still apply.
  • Optimize Mode: Seedream offers Standard and Fast options.
  • Sequential Mode: Gemini offers Disabled and Sequential; Seedream offers Disabled and Auto. Use the choices shown for your selected model.

PhotoGPT also includes Z Image Turbo and Legacy General Model (FLUX.1 [dev]), bringing the general image catalog to ten models. These use pixel-resolution presets, Steps, Seed, and Reference Image Style Strength. Z Image Turbo accepts 4 or 8 steps, with 8 as the default; the Legacy General Model uses the default 20–100 range. Seed is not exposed for the eight models in the table.

All models let you select the image count. Check the current settings and credit estimate before generating.

These settings come in handy for production quality shoots and allow for higher control over your creations. For more details or any questions related to a specific use case, feel free to hop into our Discord server and drop a message to our team.

Few Best Practices to keep in mind

  • Retrain when your source photos no longer represent your appearance. Each training or retraining run costs 1,500 credits; it is not needed for every new style.
  • Store successful/favorite generations in a separate folder in your library so your team can further build on them for future generations.

These additional options will provide you with higher control over the quality of your image as well as how your input influences your generated images.

Last updated on

On this page